diff --git a/modules/gpu/src/cuda/remap.cu b/modules/gpu/src/cuda/remap.cu
index 1ab844754b..2f5f3cf94f 100644
--- a/modules/gpu/src/cuda/remap.cu
+++ b/modules/gpu/src/cuda/remap.cu
@@ -69,7 +69,8 @@ namespace cv { namespace gpu { namespace imgproc
 
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcherStream
     {
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream)
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, 
+            const float* borderValue, cudaStream_t stream, int)
         {
             typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type work_type; 
             
@@ -87,7 +88,7 @@ namespace cv { namespace gpu { namespace imgproc
     
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcherNonStream
     {
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue)
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, int)
         {
             typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type work_type; 
             
@@ -118,10 +119,10 @@ namespace cv { namespace gpu { namespace imgproc
     }; \
     template <template <typename> class Filter, template <typename> class B> struct RemapDispatcherNonStream<Filter, B, type> \
     { \
-        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float* borderValue) \
+        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float* borderValue, int cc) \
         { \
             typedef typename TypeVec<float, VecTraits< type >::cn>::vec_type work_type; \
-            dim3 block(32, 8); \
+            dim3 block(32, cc >= 20 ? 8 : 4); \
             dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \
             TextureBinder texHandler(&tex_remap_ ## type , src); \
             tex_remap_ ## type ##_reader texSrc; \
@@ -135,7 +136,7 @@ namespace cv { namespace gpu { namespace imgproc
     }; \
     template <template <typename> class Filter> struct RemapDispatcherNonStream<Filter, BrdReplicate, type> \
     { \
-        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float*) \
+        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float*, int) \
         { \
             dim3 block(32, 8); \
             dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \
@@ -176,18 +177,21 @@ namespace cv { namespace gpu { namespace imgproc
 
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcher
     { 
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream)
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, 
+            const float* borderValue, cudaStream_t stream, int cc)
         {
             if (stream == 0)
-                RemapDispatcherNonStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue);
+                RemapDispatcherNonStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, cc);
             else
-                RemapDispatcherStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, stream);
+                RemapDispatcherStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, stream, cc);
         }
     };
 
-    template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream)
+    template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, 
+        int borderMode, const float* borderValue, cudaStream_t stream, int cc)
     {
-        typedef void (*caller_t)(const DevMem2D_<T>& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream);
+        typedef void (*caller_t)(const DevMem2D_<T>& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D_<T>& dst, 
+            const float* borderValue, cudaStream_t stream, int cc);
 
         static const caller_t callers[3][5] = 
         {
@@ -214,36 +218,36 @@ namespace cv { namespace gpu { namespace imgproc
             }
         };
 
-        callers[interpolation][borderMode](static_cast< DevMem2D_<T> >(src), xmap, ymap, static_cast< DevMem2D_<T> >(dst), borderValue, stream);
+        callers[interpolation][borderMode](static_cast< DevMem2D_<T> >(src), xmap, ymap, static_cast< DevMem2D_<T> >(dst), borderValue, stream, cc);
     }
 
-    template void remap_gpu<uchar >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<uchar2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<uchar3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<uchar4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void remap_gpu<uchar >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<uchar2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<uchar3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<uchar4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     
-    //template void remap_gpu<schar>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<char2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<char3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<char4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    //template void remap_gpu<schar>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<char2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<char3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<char4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     
-    template void remap_gpu<ushort >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<ushort2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<ushort3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<ushort4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void remap_gpu<ushort >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<ushort2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<ushort3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<ushort4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     
-    template void remap_gpu<short >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<short2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<short3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<short4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void remap_gpu<short >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<short2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<short3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<short4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     
-    //template void remap_gpu<int >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<int2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<int3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<int4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    //template void remap_gpu<int >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<int2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<int3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<int4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     
-    template void remap_gpu<float >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    //template void remap_gpu<float2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<float3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
-    template void remap_gpu<float4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    template void remap_gpu<float >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    //template void remap_gpu<float2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<float3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
+    template void remap_gpu<float4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
 }}}
diff --git a/modules/gpu/src/imgproc.cpp b/modules/gpu/src/imgproc.cpp
index 528ac65794..3b47447a12 100644
--- a/modules/gpu/src/imgproc.cpp
+++ b/modules/gpu/src/imgproc.cpp
@@ -110,14 +110,15 @@ void cv::gpu::CannyBuf::release() { throw_nogpu(); }
 namespace cv { namespace gpu {  namespace imgproc
 {
     template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, 
-                                         int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+                                         int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);
 }}}
 
 void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode, const Scalar& borderValue, Stream& stream)
 {
     using namespace cv::gpu::imgproc;
 
-    typedef void (*caller_t)(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
+    typedef void (*caller_t)(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, 
+        int borderMode, const float* borderValue, cudaStream_t stream, int cc);
     static const caller_t callers[6][4] = 
     {
         {remap_gpu<uchar>, 0/*remap_gpu<uchar2>*/, remap_gpu<uchar3>, remap_gpu<uchar4>},
@@ -145,7 +146,10 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
     Scalar_<float> borderValueFloat;
     borderValueFloat = borderValue;
 
-    func(src, xmap, ymap, dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream));
+    DeviceInfo info;
+    int cc = info.majorVersion() * 10 + info.minorVersion();
+
+    func(src, xmap, ymap, dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream), cc);
 }
 
 ////////////////////////////////////////////////////////////////////////