[GSoC] New universal intrinsic backend for RVV
* Add new rvv backend (partially implemented).
* Modify the framework of Universal Intrinsic.
* Add CV_SIMD macro guards to current UI code.
* Use vlanes() instead of nlanes.
* Modify the UI test.
* Enable the new RVV (scalable) backend.
* Remove whitespace.
* Rename and some others modify.
* Update intrin.hpp but still not work on AVX/SSE
* Update conditional compilation macros.
* Use static variable for vlanes.
* Use max_nlanes for array defining.
It's not clear how ranges argument should be used in the overload of
calcHist that accepts std::vector. The main overload uses array of
arrays there, while std::vector overload uses a plain array. The code
interprets the vector as a flattened array and rebuilds array of arrays
from it. This is not obvious interpretation, so documentation has been
added to explain the expected usage.
Replaced sprintf with safer snprintf
* Straightforward replacement of sprintf with safer snprintf
* Trickier replacement of sprintf with safer snprintf
Some functions were changed to take another parameter: the size of the buffer, so that they can pass that size on to snprintf.
Fixed out-of-bounds read in parallel version of ippGaussianBlur()
* Fixed out-of-memory read in parallel version of ippGaussianBlur()
* Fixed check
* Revert changes in CMakeLists.txt
* better accuracy of _rotatedRectangleIntersection
instead of just migrating to double-precision (which would work), some computations are scaled by a factor that depends on the length of the smallest vectors.
There is a better accuracy even with floats, so this is certainly better for very sensitive cases
* Update intersection.cpp
use L2SQR norm to tune the numeric scale
* Update intersection.cpp
adapt samePointEps with L2 norm
* Update intersection.cpp
move comment
* Update intersection.cpp
fix wrong numericalScalingFactor usage
* added tests
* fixed warnings returned by buildbot
* modifications suggested by reviewer
renaming numericalScaleFctor to normalizationScale
refactor some computations
more "const"
* modifications as suggested by reviewer
Optimize cv::applyColorMap() for simple case
* Optimize cv::applyColorMap() for simple case
PR for 21640
For regular cv::Mat CV_8UC1 src, applying the colormap is simpler than calling the cv::LUT() mechanism.
* add support for src as CV_8UC3
src as CV_8UC3 is handled with a BGR2GRAY conversion, the same optimized code being used afterwards
* code style
rely on cv::Mat.ptr() to index data
* Move new implementation to ColorMap::operator()
Changes as suggested by reviewer
* style
improvements suggsted by reviewer
* typo
* tune parallel work
* better usage of parallel_for_
use nstripes parameter of parallel_for_
assume _lut is continuous to bring faster pixel indexing
optimize src/dst access by contiguous rows of pixels
do not locally copy the LUT any more, it is no more relevant with the new optimizations
Fixed threshold(THRESH_TOZERO) at imgproc(IPP)
* Fixed#16085: imgproc(IPP): wrong result from threshold(THRESH_TOZERO)
* 1. Added test cases with float where all bits of mantissa equal 1, min and max float as inputs
2. Used nextafterf instead of cast to hex
* Used float value in test instead of hex and casts
* Changed input value in test
When computing:
t1 = (bayer[1] + bayer[bayer_step] + bayer[bayer_step+2] + bayer[bayer_step*2+1])*G2Y;
there is a T (unsigned short or char) multiplied by an int which can overflow.
Then again, it is stored to t1 which is unsigned so the overflow disappears.
Keeping all unsigned is safer.
* Fix integer overflow in cv::Luv2RGBinteger::process.
For LL=49, uu=205, vv=23, we end up with x=7373056 and y=458
which overflows y*x.
* imgproc(test): adjust test parameters to cover SIMD code
In case of very small negative h (e.g. -1e-40), with the current implementation,
you will go through the first condition and end up with h = 6.f, and will miss
the second condition.
* Add RowVec_8u32f
* Fix build errors in Linux x64 Debug and armeabi-v7a
* Reformat code to make it more clean and conventional
* Optimise with vx_load_expand_q()
different paddings in cvtColorTwoPlane() for biplane YUV420
* Different paddings support in cvtColorTwoPlane() for biplane YUV420
* Build fix for dispatch case.
* Resoted old behaviour for y.step==uv.step to exclude perf regressions.
Co-authored-by: amir.tulegenov <amir.tulegenov@xperience.ai>
Co-authored-by: Alexander Smorkalov <alexander.smorkalov@xperience.ai>
Update rotatedRectangleIntersection function to calculate near to origin
* Change type used in points function from RotatedRect
In the function that sets the points of a RotatedRect, the types
should be double in order to keep the precision when dealing with
RotatedRects that are defined far from the origin.
This commit solves the problem in some assertions from
rotatedRectangleIntersection when dealing with rectangles far from
origin.
* added proper type casts
* Update rotatedRectangleIntersection function to calculate near to origin
This commit changes the rotatedRectangleIntersection function in order
to calculate the intersection of two rectangles considering that they
are shifted near the coordinates origin (0, 0).
This commit solves the problem in some assertions from
rotatedRectangleIntersection when dealing with rectangles far from
origin.
* Revert type changes in types.cpp and adequate code to c++98
* Revert unnecessary casts on types.cpp
Co-authored-by: Vadim Pisarevsky <vadim.pisarevsky@gmail.com>
* Add Neon optimised RGB2Lab conversion
* Fix compile errors, change lambda to macro
* Change NEON optimised RGB2Lab to just use HAL
* Change [] to v_extract_n in RGB2Lab
* RGB2LAB Code quality, change to nlane agnostic
* Change RGB2Lab to use function rather than macro
* Remove whitespace
Co-authored-by: Francesco Petrogalli <25690309+fpetrogalli@users.noreply.github.com>
Also bring perf_imgproc CornerMinEigenVal accuracy requirements in line with
the test_imgproc accuracy requirements on that test and fix indentation on
the latter.
Partially addresses issue #9821
The MinEigenVal path through the corner.cl kernel makes use of native_sqrt,
a math builtin function which has implementation defined accuracy.
Partially addresses issue #9821
* goodFeaturesToTrack returns also corner value
(cherry picked from commit 4a8f06755c)
* Added response to GFTT Detector keypoints
(cherry picked from commit b88fb40c6e)
* Moved corner values to another optional variable to preserve backward compatibility
(cherry picked from commit 6137383d32)
* Removed corners valus from perf tests and better unit tests for corners values
(cherry picked from commit f3d0ef21a7)
* Fixed detector gftt call
(cherry picked from commit be2975553b)
* Restored test_cornerEigenValsVecs
(cherry picked from commit ea3e11811f)
* scaling fixed;
mineigen calculation rolled back;
gftt function overload added (with quality parameter);
perf tests were added for the new api function;
external bindings were added for the function (with different alias);
fixed issues with composition of the output array of the new function (e.g. as requested in comments) ;
added sanity checks in the perf tests;
removed C API changes.
* minor change to GFTTDetector::detect
* substitute ts->printf with EXPECT_LE
* avoid re-allocations
Co-authored-by: Anas <anas.el.amraoui@live.com>
Co-authored-by: amir.tulegenov <amir.tulegenov@xperience.ai>
Return accumulator value in HoughLines algorithm
* try to solve #17050
use cv_wrap_as
add python test
parameters
* review
* move wrapper to imgproc/bindings.hpp
Bit exact gaussian blur for 16bit unsigned int
* bit-exact gaussian kernel for CV_16U
* SIMD optimization
* template GaussianBlurFixedPoint
* remove template specialization
* simd support for h3N121 uint16
* test for u16 gaussian blur
* remove unnecessary comments
* fix return type of raw()
* add typedef of native internal type in fixedpoint
* update return type of raw()
Bit-exact Nearest Neighbor Resizing
* bit exact resizeNN
* change the value of method enum
* add bitexact-nn to ResizeExactTest
* test to compare with non-exact version
* add perf for bit-exact resizenn
* use cvFloor-equivalent
* 1/3 scaling is not stable for floating calculation
* stricter test
* bugfix: broken data in case of 6 or 12bytes elements
* bugfix: broken data in default pix_size
* stricter threshold
* use raw() for floor
* use double instead of int
* follow code reviews
* fewer cases in perf test
* center pixel convention
changed OpenCV license from BSD to Apache 2 license
* as discussed and announced earlier, changed OpenCV license from BSD to Apache 2. Many files still contain old-style copyrights though
* changed wording a bit; preserve the original OpenCV BSD license
Clarify component statistics documentation
* Change ConnectedComponentsTypes documentation
Change from "algorithm output formats" to "statistics" because it specifies types of statistics, not formats.
* Documentation: clarify component statistics
Explain that ConnectedComponentTypes selects a statistic.
Objc binding
* Initial work on Objective-C wrapper
* Objective-C generator script; update manually generated wrappers
* Add Mat tests
* Core Tests
* Imgproc wrapper generation and tests
* Fixes for Imgcodecs wrapper
* Miscellaneous fixes. Swift build support
* Objective-C wrapper build/install
* Add Swift wrappers for videoio/objdetect/feature2d
* Framework build;iOS support
* Fix toArray functions;Use enum types whenever possible
* Use enum types where possible;prepare test build
* Update test
* Add test runner scripts for iOS and macOS
* Add test scripts and samples
* Build fixes
* Fix build (cmake 3.17.x compatibility)
* Fix warnings
* Fix enum name conflicting handling
* Add support for document generation with Jazzy
* Swift/Native fast accessor functions
* Add Objective-C wrapper for calib3d, dnn, ml, photo and video modules
* Remove IntOut/FloatOut/DoubleOut classes
* Fix iOS default test platform value
* Fix samples
* Revert default framework name to opencv2
* Add converter util functions
* Fix failing test
* Fix whitespace
* Add handling for deprecated methods;fix warnings;define __OPENCV_BUILD
* Suppress cmake warnings
* Reduce severity of "jazzy not found" log message
* Fix incorrect #include of compatibility header in ios.h
* Use explicit returns in subscript/get implementation
* Reduce minimum required cmake version to 3.15 for Objective-C/Swift binding
* improved fitEllipse and fitEllipseDirect accuracy in singular or close-to-singular cases (see issue #9923)
* scale points using double precision
* added normalization to fitEllipseAMS as well; fixed Java test case by raising the tolerance (it's unclear what is the correct result in this case).
* improved point perturbation a bit. make the code a little bit more clear
* trying to fix Java fitEllipseTest by slightly raising the tolerance threshold
* synchronized C++ version of Java's fitEllipse test
* removed trailing whitespaces
* imgproc: templmatch: Add support for mask for all methods
Add support for masked template matching. Fix/scrub old implementation
for masked matching, as it did partly not even really do a meaningful
masking, and only supported limited template matching methods.
Add documentation including formulas for masked matching.
* imgproc: test: Add tests for masked template matching
Test accuracy by comparing to naive implementation for one point.
Test compatibility/correctness by comparing results without mask and
with all ones mask.
All tests are done for all methods, all supported depths, and for 1 and
3 channels.
* imgproc: test: templmatch: Add test for crossCorr
Add a test for the crossCorr function in templmatch.cpp. crossCorr() had
to be added to exported functions to be testable.
This test can maybe help to identify the problem with template matching
on MacOSX.
* fix: Fixed wrong evaluations of the MatExpr on Clang
* fix: removed crossCorr from public interface.
If it should be exported, it should be done as separate PR.
Co-authored-by: Vadim Levin <vadim.levin@xperience.ai>
* LineVirtualIterator
Proposal of LineVirtualIterator, an alternative to "LineIterator not attached to any mat".
This is basically the same implementation, replacing the address difference by a single "offset" variable. elemsize becomes irrelevant and considered to be 1. "step" is thus equal to size.width since no stride is expected.
* Update drawing.cpp
fixed warning
* improvement of LineVirtualIterator
instead of being too conservative, the new implementation gets rid of "offset/step" and only keeps a "Point currentPos" up to date.
left_to_right is renamed to forceLeftToRight as suggested (even for the old LineIterator)
assert() replaced by CV_Assert() (even for the old LineIterator)
* fixed implementation
+fixed last commit so that LineVirtualIterator gives at least the same results as LineIterator
+added a new constructor that does not require any Size, so that no clipping is done and iteration occurs from pt1 to pt2. This is done by adding a spatial offset to pt1 and pt2 so that the same implementation is used, the size being in that case the spatial size between pt1 and pt2
* Update imgproc.hpp
fixed warnings
* Update drawing.cpp
fixed whitespace
* Update drawing.cpp
trailing whitespace
* Update imgproc.hpp
+added a new constructor that takes a Rect rather than a Size. It computes the line pt1->pt2 that clips that rect.
Yet again, this is still based on the same implementation, thanks to the Size and the currentPosOffset that can artifically consider the origin of the rect at (0,0)
* revert changes
revert changes on original LineIterator implementation, that will be superseded by the new LineVirtualIterator anyway
* added test of LineVirtualIterator
* More tests
* refactoring
Use C++11 chained constructors
Improved code style
* improve test
Added offset as random test data.
* fixed order of initialization
* merged LineIterator and VirtualLineIterator
* merged LineIterator & VirtualLineIterator
* merged LineIterator & VirtualLineIterator
* merged LineIterator & VirtualLineIterator
* made LineIterator::operator ++() more efficient
added one perfectly predictable check; in theory, since ptmode is set in the end of the constructor in the header file, the compiler can figure out that it's always true/false and eliminate the check from the inline `LineIterator::operator++()` completely
* optimized Line() function
in the most common case (CV_8UC3) eliminated the check from the loop
Co-authored-by: Vadim Pisarevsky <vadim.pisarevsky@gmail.com>
