diff --git a/CMakeLists.txt b/CMakeLists.txt
index c87d6fbf91..f9baf672eb 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -197,7 +197,6 @@ OCV_OPTION(ENABLE_SSE41               "Enable SSE4.1 instructions"
 OCV_OPTION(ENABLE_SSE42               "Enable SSE4.2 instructions"                               OFF  IF (CMAKE_COMPILER_IS_GNUCXX AND (X86 OR X86_64)) )
 OCV_OPTION(ENABLE_NOISY_WARNINGS      "Show all warnings even if they are too noisy"             OFF )
 OCV_OPTION(OPENCV_WARNINGS_ARE_ERRORS "Treat warnings as errors"                                 OFF )
-OCV_OPTION(ENABLE_MULTI_PROCESSOR_COMPILATION "Enabling multi-processory compilation"            OFF IF MSVC)
 
 
 # uncategorized options
diff --git a/cmake/OpenCVCompilerOptions.cmake b/cmake/OpenCVCompilerOptions.cmake
index b3d71c8c8d..1913527451 100644
--- a/cmake/OpenCVCompilerOptions.cmake
+++ b/cmake/OpenCVCompilerOptions.cmake
@@ -282,9 +282,4 @@ if(MSVC)
   if(NOT ENABLE_NOISY_WARNINGS)
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /wd4251") #class 'std::XXX' needs to have dll-interface to be used by clients of YYY
   endif()
-endif()
-
-
-if (MSVC AND ENABLE_MULTI_PROCESSOR_COMPILATION)
-   SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /MP")
-endif()
+endif()
\ No newline at end of file
diff --git a/modules/gpu/doc/image_processing.rst b/modules/gpu/doc/image_processing.rst
index 858b707bae..c8fd7491e7 100644
--- a/modules/gpu/doc/image_processing.rst
+++ b/modules/gpu/doc/image_processing.rst
@@ -818,9 +818,57 @@ Performs linear blending of two images.
     :param result: Destination image.
 
     :param stream: Stream for the asynchronous version.
+    
+    
+gpu::bilateralFilter
+-------------------
+Performs bilateral filtering of passed image
+
+.. ocv:function:: void gpu::bilateralFilter(const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode, Stream& stream = Stream::Null());
+    
+    :param src: Source image. Supports only (channles != 2 && depth() != CV_8S && depth() != CV_32S && depth() != CV_64F).
+
+    :param dst: Destination imagwe.
+
+    :param kernel_size: Kernel window size.
+
+    :param sigma_color: Filter sigma in the color space. 
+    
+    :param sigma_spatial:  Filter sigma in the coordinate space. 
+
+    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param stream: Stream for the asynchronous version.
 
+.. seealso::
+
+    :ocv:func:`bilateralFilter`,
+    
+    
+gpu::nonLocalMeans
+-------------------
+Performs pure non local means denoising without any simplification, and thus it is not fast.
+
+.. ocv:function:: void nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_widow_size = 11, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null());
+    
+    :param src: Source image. Supports only CV_8UC1, CV_8UC3.
+
+    :param dst: Destination imagwe.
+
+    :param h: Filter sigma regulating filter strength for color. 
+    
+    :param search_widow_size: Size of search window.
+
+    :param block_size: Size of block used for computing weights. 
+        
+    :param borderMode:  Border type. See :ocv:func:`borderInterpolate` for details. ``BORDER_REFLECT101`` , ``BORDER_REPLICATE`` , ``BORDER_CONSTANT`` , ``BORDER_REFLECT`` and ``BORDER_WRAP`` are supported for now.
+
+    :param stream: Stream for the asynchronous version.
 
+.. seealso::
 
+    :ocv:func:`fastNlMeansDenoising`
+    
 gpu::alphaComp
 -------------------
 Composites two images using alpha opacity values contained in each image.
diff --git a/modules/gpu/include/opencv2/gpu/gpu.hpp b/modules/gpu/include/opencv2/gpu/gpu.hpp
index c2fcc31a2e..2faa1751d4 100644
--- a/modules/gpu/include/opencv2/gpu/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu/gpu.hpp
@@ -769,6 +769,14 @@ CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::N
 CV_EXPORTS void blendLinear(const GpuMat& img1, const GpuMat& img2, const GpuMat& weights1, const GpuMat& weights2,
                             GpuMat& result, Stream& stream = Stream::Null());
 
+//! Performa bilateral filtering of passsed image
+CV_EXPORTS void bilateralFilter(const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial, 
+                                int borderMode = BORDER_DEFAULT, Stream& stream = Stream::Null());
+
+//! Brute force non-local means algorith (slow but universal)
+CV_EXPORTS void nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, 
+                              int search_widow_size = 11, int block_size = 7, int borderMode = BORDER_DEFAULT, Stream& s = Stream::Null());
+
 
 struct CV_EXPORTS CannyBuf;
 
diff --git a/modules/gpu/perf/perf_core.cpp b/modules/gpu/perf/perf_core.cpp
index b638fbf4ad..915a9a20b4 100644
--- a/modules/gpu/perf/perf_core.cpp
+++ b/modules/gpu/perf/perf_core.cpp
@@ -882,7 +882,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseAndMat, Combine(GPU_TYPICAL_MAT_SIZES, Values(
 //////////////////////////////////////////////////////////////////////
 // BitwiseAndScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseAndScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseAndScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), GPU_CHANNELS_1_3_4))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -963,7 +963,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseOrMat, Combine(GPU_TYPICAL_MAT_SIZES, Values(C
 //////////////////////////////////////////////////////////////////////
 // BitwiseOrScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseOrScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseOrScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), GPU_CHANNELS_1_3_4))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -1044,7 +1044,7 @@ PERF_TEST_P(Sz_Depth, Core_BitwiseXorMat, Combine(GPU_TYPICAL_MAT_SIZES, Values(
 //////////////////////////////////////////////////////////////////////
 // BitwiseXorScalar
 
-PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseXorScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseXorScalar, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), GPU_CHANNELS_1_3_4))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -1085,7 +1085,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_BitwiseXorScalar, Combine(GPU_TYPICAL_MAT_SIZES, V
 //////////////////////////////////////////////////////////////////////
 // RShift
 
-PERF_TEST_P(Sz_Depth_Cn, Core_RShift, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, Core_RShift, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), GPU_CHANNELS_1_3_4))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -1119,7 +1119,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_RShift, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8
 //////////////////////////////////////////////////////////////////////
 // LShift
 
-PERF_TEST_P(Sz_Depth_Cn, Core_LShift, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, Core_LShift, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32S), GPU_CHANNELS_1_3_4))
 {
     const cv::Size size = GET_PARAM(0);
     const int depth = GET_PARAM(1);
@@ -1461,7 +1461,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Code, cv::Size, MatDepth, int, FlipCode);
 PERF_TEST_P(Sz_Depth_Cn_Code, Core_Flip, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     ALL_FLIP_CODES))
 {
     cv::Size size = GET_PARAM(0);
@@ -1973,7 +1973,7 @@ PERF_TEST_P(Sz_Norm, Core_NormDiff, Combine(
 PERF_TEST_P(Sz_Depth_Cn, Core_Sum, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4)))
+    GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -2015,7 +2015,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_Sum, Combine(
 PERF_TEST_P(Sz_Depth_Cn, Core_SumAbs, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4)))
+    GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -2052,7 +2052,7 @@ PERF_TEST_P(Sz_Depth_Cn, Core_SumAbs, Combine(
 PERF_TEST_P(Sz_Depth_Cn, Core_SumSqr, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values<MatDepth>(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4)))
+    GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
diff --git a/modules/gpu/perf/perf_denoising.cpp b/modules/gpu/perf/perf_denoising.cpp
new file mode 100644
index 0000000000..ee76b55943
--- /dev/null
+++ b/modules/gpu/perf/perf_denoising.cpp
@@ -0,0 +1,98 @@
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace testing;
+
+
+//////////////////////////////////////////////////////////////////////
+// BilateralFilter
+
+DEF_PARAM_TEST(Sz_Depth_Cn_KernelSz, cv::Size, MatDepth , int, int);
+
+PERF_TEST_P(Sz_Depth_Cn_KernelSz, Denoising_BilateralFilter, 
+            Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F), GPU_CHANNELS_1_3_4, Values(3, 5, 9)))
+{
+    declare.time(30.0);
+
+    cv::Size size = GET_PARAM(0);
+    int depth = GET_PARAM(1);
+    int channels = GET_PARAM(2);
+    int kernel_size = GET_PARAM(3);
+
+    float sigma_color = 7;
+    float sigma_spatial = 5;
+    int borderMode = cv::BORDER_REFLECT101;
+
+    int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    fillRandom(src);
+
+     if (runOnGpu)
+    {
+        cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_dst;
+
+        cv::gpu::bilateralFilter(d_src, d_dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+
+        TEST_CYCLE()
+        {
+            cv::gpu::bilateralFilter(d_src, d_dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+        }
+    }
+    else
+    {
+        cv::Mat dst;
+
+        cv::bilateralFilter(src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+
+        TEST_CYCLE()
+        {
+            cv::bilateralFilter(src, dst, kernel_size, sigma_color, sigma_spatial, borderMode);
+        }
+    }
+}
+
+
+//////////////////////////////////////////////////////////////////////
+// nonLocalMeans
+
+DEF_PARAM_TEST(Sz_Depth_Cn_WinSz_BlockSz, cv::Size, MatDepth , int, int, int);
+
+PERF_TEST_P(Sz_Depth_Cn_WinSz_BlockSz, Denoising_NonLocalMeans, 
+            Combine(GPU_TYPICAL_MAT_SIZES, Values<MatDepth>(CV_8U), Values(1), Values(21), Values(5, 7)))
+{
+    declare.time(30.0);
+
+    cv::Size size = GET_PARAM(0);
+    int depth = GET_PARAM(1);
+    int channels = GET_PARAM(2);
+    
+    int search_widow_size = GET_PARAM(3);
+    int block_size = GET_PARAM(4);
+
+    float h = 10;
+    int borderMode = cv::BORDER_REFLECT101;
+    
+    int type = CV_MAKE_TYPE(depth, channels);
+
+    cv::Mat src(size, type);
+    fillRandom(src);
+
+    if (runOnGpu)
+    {
+        cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_dst;
+
+        cv::gpu::nonLocalMeans(d_src, d_dst, h, search_widow_size, block_size, borderMode);
+
+        TEST_CYCLE()
+        {
+            cv::gpu::nonLocalMeans(d_src, d_dst, h, search_widow_size, block_size, borderMode);
+        }
+    }
+    else
+    {
+        FAIL();
+    }
+}
\ No newline at end of file
diff --git a/modules/gpu/perf/perf_imgproc.cpp b/modules/gpu/perf/perf_imgproc.cpp
index 80d4af54df..761510da35 100644
--- a/modules/gpu/perf/perf_imgproc.cpp
+++ b/modules/gpu/perf/perf_imgproc.cpp
@@ -54,7 +54,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border_Mode, cv::Size, MatDepth, int, Interpola
 PERF_TEST_P(Sz_Depth_Cn_Inter_Border_Mode, ImgProc_Remap, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
     ALL_BORDER_MODES,
     ALL_REMAP_MODES))
@@ -113,7 +113,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Scale, cv::Size, MatDepth, int, Interpolation,
 PERF_TEST_P(Sz_Depth_Cn_Inter_Scale, ImgProc_Resize, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     ALL_INTERPOLATIONS,
     Values(0.5, 0.3, 2.0)))
 {
@@ -163,7 +163,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Scale, cv::Size, MatDepth, int, double);
 PERF_TEST_P(Sz_Depth_Cn_Scale, ImgProc_ResizeArea, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(0.2, 0.1, 0.05)))
 {
     declare.time(1.0);
@@ -212,7 +212,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Inter_Border, cv::Size, MatDepth, int, Interpolation,
 PERF_TEST_P(Sz_Depth_Cn_Inter_Border, ImgProc_WarpAffine, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
     ALL_BORDER_MODES))
 {
@@ -265,7 +265,7 @@ PERF_TEST_P(Sz_Depth_Cn_Inter_Border, ImgProc_WarpAffine, Combine(
 PERF_TEST_P(Sz_Depth_Cn_Inter_Border, ImgProc_WarpPerspective, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC)),
     ALL_BORDER_MODES))
 {
@@ -321,7 +321,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Border, cv::Size, MatDepth, int, BorderMode);
 PERF_TEST_P(Sz_Depth_Cn_Border, ImgProc_CopyMakeBorder, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     ALL_BORDER_MODES))
 {
     cv::Size size = GET_PARAM(0);
@@ -789,7 +789,7 @@ PERF_TEST_P(Image, ImgProc_MeanShiftSegmentation, Values<string>("gpu/meanshift/
 //////////////////////////////////////////////////////////////////////
 // BlendLinear
 
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_BlendLinear, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_32F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, ImgProc_BlendLinear, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_32F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -887,7 +887,7 @@ DEF_PARAM_TEST(Sz_TemplateSz_Cn_Method, cv::Size, cv::Size, int, TemplateMethod)
 PERF_TEST_P(Sz_TemplateSz_Cn_Method, ImgProc_MatchTemplate8U, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     ALL_TEMPLATE_METHODS))
 {
     cv::Size size = GET_PARAM(0);
@@ -933,7 +933,7 @@ PERF_TEST_P(Sz_TemplateSz_Cn_Method, ImgProc_MatchTemplate8U, Combine(
 PERF_TEST_P(Sz_TemplateSz_Cn_Method, ImgProc_MatchTemplate32F, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(cv::Size(5, 5), cv::Size(16, 16), cv::Size(30, 30)),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(TemplateMethod(cv::TM_SQDIFF), TemplateMethod(cv::TM_CCORR))))
 {
     cv::Size size = GET_PARAM(0);
@@ -1287,7 +1287,7 @@ DEF_PARAM_TEST(Sz_Depth_Cn_Inter, cv::Size, MatDepth, int, Interpolation);
 PERF_TEST_P(Sz_Depth_Cn_Inter, ImgProc_Rotate, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4),
+    GPU_CHANNELS_1_3_4,
     Values(Interpolation(cv::INTER_NEAREST), Interpolation(cv::INTER_LINEAR), Interpolation(cv::INTER_CUBIC))))
 {
     cv::Size size = GET_PARAM(0);
@@ -1324,7 +1324,7 @@ PERF_TEST_P(Sz_Depth_Cn_Inter, ImgProc_Rotate, Combine(
 PERF_TEST_P(Sz_Depth_Cn, ImgProc_PyrDown, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4)))
+    GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -1366,7 +1366,7 @@ PERF_TEST_P(Sz_Depth_Cn, ImgProc_PyrDown, Combine(
 PERF_TEST_P(Sz_Depth_Cn, ImgProc_PyrUp, Combine(
     GPU_TYPICAL_MAT_SIZES,
     Values(CV_8U, CV_16U, CV_32F),
-    Values(1, 3, 4)))
+    GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -1540,7 +1540,7 @@ PERF_TEST_P(Sz_Type_Op, ImgProc_AlphaComp, Combine(GPU_TYPICAL_MAT_SIZES, Values
 //////////////////////////////////////////////////////////////////////
 // ImagePyramidBuild
 
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidBuild, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidBuild, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -1573,7 +1573,7 @@ PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidBuild, Combine(GPU_TYPICAL_MAT_SIZE
 //////////////////////////////////////////////////////////////////////
 // ImagePyramidGetLayer
 
-PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidGetLayer, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, ImgProc_ImagePyramidGetLayer, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
diff --git a/modules/gpu/perf/perf_matop.cpp b/modules/gpu/perf/perf_matop.cpp
index cdae962f23..83e27875a2 100644
--- a/modules/gpu/perf/perf_matop.cpp
+++ b/modules/gpu/perf/perf_matop.cpp
@@ -8,7 +8,7 @@ namespace {
 //////////////////////////////////////////////////////////////////////
 // SetTo
 
-PERF_TEST_P(Sz_Depth_Cn, MatOp_SetTo, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, MatOp_SetTo, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -45,7 +45,7 @@ PERF_TEST_P(Sz_Depth_Cn, MatOp_SetTo, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8
 //////////////////////////////////////////////////////////////////////
 // SetToMasked
 
-PERF_TEST_P(Sz_Depth_Cn, MatOp_SetToMasked, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, MatOp_SetToMasked, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
@@ -87,7 +87,7 @@ PERF_TEST_P(Sz_Depth_Cn, MatOp_SetToMasked, Combine(GPU_TYPICAL_MAT_SIZES, Value
 //////////////////////////////////////////////////////////////////////
 // CopyToMasked
 
-PERF_TEST_P(Sz_Depth_Cn, MatOp_CopyToMasked, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), Values(1, 3, 4)))
+PERF_TEST_P(Sz_Depth_Cn, MatOp_CopyToMasked, Combine(GPU_TYPICAL_MAT_SIZES, Values(CV_8U, CV_16U, CV_32F, CV_64F), GPU_CHANNELS_1_3_4))
 {
     cv::Size size = GET_PARAM(0);
     int depth = GET_PARAM(1);
diff --git a/modules/gpu/perf/perf_video.cpp b/modules/gpu/perf/perf_video.cpp
index aa802fbf54..7a64034691 100644
--- a/modules/gpu/perf/perf_video.cpp
+++ b/modules/gpu/perf/perf_video.cpp
@@ -423,7 +423,7 @@ PERF_TEST_P(Video, Video_FGDStatModel, Values("gpu/video/768x576.avi", "gpu/vide
 
 DEF_PARAM_TEST(Video_Cn_LearningRate, string, int, double);
 
-PERF_TEST_P(Video_Cn_LearningRate, Video_MOG, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), Values(1, 3, 4), Values(0.0, 0.01)))
+PERF_TEST_P(Video_Cn_LearningRate, Video_MOG, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), GPU_CHANNELS_1_3_4, Values(0.0, 0.01)))
 {
     string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
     int cn = GET_PARAM(1);
@@ -511,7 +511,7 @@ PERF_TEST_P(Video_Cn_LearningRate, Video_MOG, Combine(Values("gpu/video/768x576.
 
 DEF_PARAM_TEST(Video_Cn, string, int);
 
-PERF_TEST_P(Video_Cn, Video_MOG2, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), Values(1, 3, 4)))
+PERF_TEST_P(Video_Cn, Video_MOG2, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), GPU_CHANNELS_1_3_4))
 {
     string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
     int cn = GET_PARAM(1);
@@ -596,7 +596,7 @@ PERF_TEST_P(Video_Cn, Video_MOG2, Combine(Values("gpu/video/768x576.avi", "gpu/v
 //////////////////////////////////////////////////////
 // MOG2GetBackgroundImage
 
-PERF_TEST_P(Video_Cn, Video_MOG2GetBackgroundImage, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), Values(1, 3, 4)))
+PERF_TEST_P(Video_Cn, Video_MOG2GetBackgroundImage, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), GPU_CHANNELS_1_3_4))
 {
     string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
     int cn = GET_PARAM(1);
@@ -676,7 +676,7 @@ PERF_TEST_P(Video_Cn, Video_MOG2GetBackgroundImage, Combine(Values("gpu/video/76
 //////////////////////////////////////////////////////
 // VIBE
 
-PERF_TEST_P(Video_Cn, Video_VIBE, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), Values(1, 3, 4)))
+PERF_TEST_P(Video_Cn, Video_VIBE, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), GPU_CHANNELS_1_3_4))
 {
     string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
     int cn = GET_PARAM(1);
@@ -739,7 +739,7 @@ PERF_TEST_P(Video_Cn, Video_VIBE, Combine(Values("gpu/video/768x576.avi", "gpu/v
 
 DEF_PARAM_TEST(Video_Cn_MaxFeatures, string, int, int);
 
-PERF_TEST_P(Video_Cn_MaxFeatures, Video_GMG, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), Values(1, 3, 4), Values(20, 40, 60)))
+PERF_TEST_P(Video_Cn_MaxFeatures, Video_GMG, Combine(Values("gpu/video/768x576.avi", "gpu/video/1920x1080.avi"), GPU_CHANNELS_1_3_4, Values(20, 40, 60)))
 {
     std::string inputFile = perf::TestBase::getDataPath(GET_PARAM(0));
     int cn = GET_PARAM(1);
diff --git a/modules/gpu/perf/utility.hpp b/modules/gpu/perf/utility.hpp
index 441d32adb1..d2e3a070f5 100644
--- a/modules/gpu/perf/utility.hpp
+++ b/modules/gpu/perf/utility.hpp
@@ -41,5 +41,6 @@ DEF_PARAM_TEST(Sz_Depth, cv::Size, MatDepth);
 DEF_PARAM_TEST(Sz_Depth_Cn, cv::Size, MatDepth, int);
 
 #define GPU_TYPICAL_MAT_SIZES testing::Values(perf::sz720p, perf::szSXGA, perf::sz1080p)
+#define GPU_CHANNELS_1_3_4 testing::Values(1, 3, 4)
 
 #endif // __OPENCV_PERF_GPU_UTILITY_HPP__
diff --git a/modules/gpu/src/cuda/bilateral_filter.cu b/modules/gpu/src/cuda/bilateral_filter.cu
index abae91d2c5..9e9135e109 100644
--- a/modules/gpu/src/cuda/bilateral_filter.cu
+++ b/modules/gpu/src/cuda/bilateral_filter.cu
@@ -12,6 +12,7 @@
 //
 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
 // Third party copyrights are property of their respective owners.
 //
 // Redistribution and use in source and binary forms, with or without modification,
@@ -28,7 +29,7 @@
 //     derived from this software without specific prior written permission.
 //
 // This software is provided by the copyright holders and contributors "as is" and
-// any express or implied warranties, including, but not limited to, the implied
+// any express or bpied warranties, including, but not limited to, the bpied
 // warranties of merchantability and fitness for a particular purpose are disclaimed.
 // In no event shall the Intel Corporation or contributors be liable for any direct,
 // indirect, incidental, special, exemplary, or consequential damages
@@ -41,186 +42,155 @@
 //M*/
 
 #include "internal_shared.hpp"
-#include "opencv2/gpu/device/limits.hpp"
 
-namespace cv { namespace gpu { namespace device
-{
-    namespace bilateral_filter
-    {
-        __constant__ float* ctable_color;
-        __constant__ float* ctable_space;
-        __constant__ size_t ctable_space_step;
+#include "opencv2/gpu/device/vec_traits.hpp"
+#include "opencv2/gpu/device/vec_math.hpp"
+#include "opencv2/gpu/device/border_interpolate.hpp"
 
-        __constant__ int cndisp;
-        __constant__ int cradius;
+using namespace cv::gpu;
 
-        __constant__ short cedge_disc;
-        __constant__ short cmax_disc;
+typedef unsigned char uchar;
+typedef unsigned short ushort;
 
-        void load_constants(float* table_color, PtrStepSzf table_space, int ndisp, int radius, short edge_disc, short max_disc)
-        {
-            cudaSafeCall( cudaMemcpyToSymbol(ctable_color, &table_color, sizeof(table_color)) );
-            cudaSafeCall( cudaMemcpyToSymbol(ctable_space, &table_space.data, sizeof(table_space.data)) );
-            size_t table_space_step = table_space.step / sizeof(float);
-            cudaSafeCall( cudaMemcpyToSymbol(ctable_space_step, &table_space_step, sizeof(size_t)) );
-
-            cudaSafeCall( cudaMemcpyToSymbol(cndisp, &ndisp, sizeof(int)) );
-            cudaSafeCall( cudaMemcpyToSymbol(cradius, &radius, sizeof(int)) );
+//////////////////////////////////////////////////////////////////////////////////
+/// Bilateral filtering
 
-            cudaSafeCall( cudaMemcpyToSymbol(cedge_disc, &edge_disc, sizeof(short)) );
-            cudaSafeCall( cudaMemcpyToSymbol(cmax_disc, &max_disc, sizeof(short)) );
-        }
-
-        template <int channels>
-        struct DistRgbMax
-        {
-            static __device__ __forceinline__ uchar calc(const uchar* a, const uchar* b)
-            {
-                uchar x = ::abs(a[0] - b[0]);
-                uchar y = ::abs(a[1] - b[1]);
-                uchar z = ::abs(a[2] - b[2]);
-                return (::max(::max(x, y), z));
-            }
-        };
+namespace cv { namespace gpu { namespace device
+{
+    namespace imgproc
+    {
+        __device__ __forceinline__ float norm_l1(const float& a)  { return ::fabs(a); }
+        __device__ __forceinline__ float norm_l1(const float2& a) { return ::fabs(a.x) + ::fabs(a.y); }
+        __device__ __forceinline__ float norm_l1(const float3& a) { return ::fabs(a.x) + ::fabs(a.y) + ::fabs(a.z); }
+        __device__ __forceinline__ float norm_l1(const float4& a) { return ::fabs(a.x) + ::fabs(a.y) + ::fabs(a.z) + ::fabs(a.w); }
 
-        template <>
-        struct DistRgbMax<1>
-        {
-            static __device__ __forceinline__ uchar calc(const uchar* a, const uchar* b)
-            {
-                return ::abs(a[0] - b[0]);
-            }
-        };
+        __device__ __forceinline__ float sqr(const float& a)  { return a * a; }
 
-        template <int channels, typename T>
-        __global__ void bilateral_filter(int t, T* disp, size_t disp_step, const uchar* img, size_t img_step, int h, int w)
+        template<typename T, typename B> 
+        __global__ void bilateral_kernel(const PtrStepSz<T> src, PtrStep<T> dst, const B b, const int ksz, const float sigma_spatial2_inv_half, const float sigma_color2_inv_half)
         {
-            const int y = blockIdx.y * blockDim.y + threadIdx.y;
-            const int x = ((blockIdx.x * blockDim.x + threadIdx.x) << 1) + ((y + t) & 1);
+            typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type value_type;
+            
+            int x = threadIdx.x + blockIdx.x * blockDim.x;
+            int y = threadIdx.y + blockIdx.y * blockDim.y;
 
-            T dp[5];
+            if (x >= src.cols || y >= src.rows)
+                return;
 
-            if (y > 0 && y < h - 1 && x > 0 && x < w - 1)
-            {
-                dp[0] = *(disp + (y  ) * disp_step + x + 0);
-                dp[1] = *(disp + (y-1) * disp_step + x + 0);
-                dp[2] = *(disp + (y  ) * disp_step + x - 1);
-                dp[3] = *(disp + (y+1) * disp_step + x + 0);
-                dp[4] = *(disp + (y  ) * disp_step + x + 1);
+            value_type center = saturate_cast<value_type>(src(y, x));
 
-                if(::abs(dp[1] - dp[0]) >= cedge_disc || ::abs(dp[2] - dp[0]) >= cedge_disc || ::abs(dp[3] - dp[0]) >= cedge_disc || ::abs(dp[4] - dp[0]) >= cedge_disc)
-                {
-                    const int ymin = ::max(0, y - cradius);
-                    const int xmin = ::max(0, x - cradius);
-                    const int ymax = ::min(h - 1, y + cradius);
-                    const int xmax = ::min(w - 1, x + cradius);
+            value_type sum1 = VecTraits<value_type>::all(0);
+            float sum2 = 0;
 
-                    float cost[] = {0.0f, 0.0f, 0.0f, 0.0f, 0.0f};
+            int r = ksz / 2;
+            float r2 = (float)(r * r);
 
-                    const uchar* ic = img + y * img_step + channels * x;
+            int tx = x - r + ksz;
+            int ty = y - r + ksz;
 
-                    for(int yi = ymin; yi <= ymax; yi++)
+            if (x - ksz/2 >=0 && y - ksz/2 >=0 && tx < src.cols && ty < src.rows)
+            {
+                for (int cy = y - r; cy < ty; ++cy)
+                    for (int cx = x - r; cx < tx; ++cx)
                     {
-                        const T* disp_y = disp + yi * disp_step;
-
-                        for(int xi = xmin; xi <= xmax; xi++)
-                        {
-                            const uchar* in = img + yi * img_step + channels * xi;
-
-                            uchar dist_rgb = DistRgbMax<channels>::calc(in, ic);
-
-                            const float weight = ctable_color[dist_rgb] * (ctable_space + ::abs(y-yi)* ctable_space_step)[::abs(x-xi)];
+                        float space2 = (x - cx) * (x - cx) + (y - cy) * (y - cy);
+                        if (space2 > r2)
+                            continue;
 
-                            const T disp_reg = disp_y[xi];
+                        value_type value = saturate_cast<value_type>(src(cy, cx));
 
-                            cost[0] += ::min(cmax_disc, ::abs(disp_reg - dp[0])) * weight;
-                            cost[1] += ::min(cmax_disc, ::abs(disp_reg - dp[1])) * weight;
-                            cost[2] += ::min(cmax_disc, ::abs(disp_reg - dp[2])) * weight;
-                            cost[3] += ::min(cmax_disc, ::abs(disp_reg - dp[3])) * weight;
-                            cost[4] += ::min(cmax_disc, ::abs(disp_reg - dp[4])) * weight;
-                        }
+                        float weight = ::exp(space2 * sigma_spatial2_inv_half + sqr(norm_l1(value - center)) * sigma_color2_inv_half);
+                        sum1 = sum1 + weight * value;
+                        sum2 = sum2 + weight;
                     }
+            }
+            else
+            {
+                for (int cy = y - r; cy < ty; ++cy)
+                    for (int cx = x - r; cx < tx; ++cx)
+                    {
+                        float space2 = (x - cx) * (x - cx) + (y - cy) * (y - cy);
+                        if (space2 > r2)
+                            continue;
 
-                    float minimum = numeric_limits<float>::max();
-                    int id = 0;
+                        value_type value = saturate_cast<value_type>(b.at(cy, cx, src.data, src.step));
 
-                    if (cost[0] < minimum)
-                    {
-                        minimum = cost[0];
-                        id = 0;
-                    }
-                    if (cost[1] < minimum)
-                    {
-                        minimum = cost[1];
-                        id = 1;
-                    }
-                    if (cost[2] < minimum)
-                    {
-                        minimum = cost[2];
-                        id = 2;
-                    }
-                    if (cost[3] < minimum)
-                    {
-                        minimum = cost[3];
-                        id = 3;
-                    }
-                    if (cost[4] < minimum)
-                    {
-                        minimum = cost[4];
-                        id = 4;
-                    }
+                        float weight = ::exp(space2 * sigma_spatial2_inv_half + sqr(norm_l1(value - center)) * sigma_color2_inv_half);
 
-                    *(disp + y * disp_step + x) = dp[id];
-                }
+                        sum1 = sum1 + weight * value;
+                        sum2 = sum2 + weight;
+                    }
             }
+            dst(y, x) = saturate_cast<T>(sum1 / sum2);
         }
 
-        template <typename T>
-        void bilateral_filter_caller(PtrStepSz<T> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream)
+        template<typename T, template <typename> class B>
+        void bilateral_caller(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, cudaStream_t stream)
         {
-            dim3 threads(32, 8, 1);
-            dim3 grid(1, 1, 1);
-            grid.x = divUp(disp.cols, threads.x << 1);
-            grid.y = divUp(disp.rows, threads.y);
+            dim3 block (32, 8);
+            dim3 grid (divUp (src.cols, block.x), divUp (src.rows, block.y));
 
-            switch (channels)
-            {
-            case 1:
-                for (int i = 0; i < iters; ++i)
-                {
-                    bilateral_filter<1><<<grid, threads, 0, stream>>>(0, disp.data, disp.step/sizeof(T), img.data, img.step, disp.rows, disp.cols);
-                    cudaSafeCall( cudaGetLastError() );
-
-                    bilateral_filter<1><<<grid, threads, 0, stream>>>(1, disp.data, disp.step/sizeof(T), img.data, img.step, disp.rows, disp.cols);
-                    cudaSafeCall( cudaGetLastError() );
-                }
-                break;
-            case 3:
-                for (int i = 0; i < iters; ++i)
-                {
-                    bilateral_filter<3><<<grid, threads, 0, stream>>>(0, disp.data, disp.step/sizeof(T), img.data, img.step, disp.rows, disp.cols);
-                    cudaSafeCall( cudaGetLastError() );
-
-                    bilateral_filter<3><<<grid, threads, 0, stream>>>(1, disp.data, disp.step/sizeof(T), img.data, img.step, disp.rows, disp.cols);
-                    cudaSafeCall( cudaGetLastError() );
-                }
-                break;
-            default:
-                cv::gpu::error("Unsupported channels count", __FILE__, __LINE__, "bilateral_filter_caller");
-            }
+            B<T> b(src.rows, src.cols);
+
+            float sigma_spatial2_inv_half = -0.5f/(sigma_spatial * sigma_spatial);
+             float sigma_color2_inv_half = -0.5f/(sigma_color * sigma_color);
+
+            cudaSafeCall( cudaFuncSetCacheConfig (bilateral_kernel<T, B<T> >, cudaFuncCachePreferL1) );
+            bilateral_kernel<<<grid, block>>>((PtrStepSz<T>)src, (PtrStepSz<T>)dst, b, kernel_size, sigma_spatial2_inv_half, sigma_color2_inv_half);
+            cudaSafeCall ( cudaGetLastError () );
 
             if (stream == 0)
                 cudaSafeCall( cudaDeviceSynchronize() );
         }
 
-        void bilateral_filter_gpu(PtrStepSzb disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream)
+        template<typename T>
+        void bilateral_filter_gpu(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float gauss_spatial_coeff, float gauss_color_coeff, int borderMode, cudaStream_t stream)
         {
-            bilateral_filter_caller(disp, img, channels, iters, stream);
-        }
+            typedef void (*caller_t)(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, cudaStream_t stream);
 
-        void bilateral_filter_gpu(PtrStepSz<short> disp, PtrStepSzb img, int channels, int iters, cudaStream_t stream)
-        {
-            bilateral_filter_caller(disp, img, channels, iters, stream);
+            static caller_t funcs[] = 
+            {
+                bilateral_caller<T, BrdReflect101>,
+                bilateral_caller<T, BrdReplicate>,
+                bilateral_caller<T, BrdConstant>,
+                bilateral_caller<T, BrdReflect>,
+                bilateral_caller<T, BrdWrap>,
+            };
+            funcs[borderMode](src, dst, kernel_size, gauss_spatial_coeff, gauss_color_coeff, stream);
         }
-    } // namespace bilateral_filter
-}}} // namespace cv { namespace gpu { namespace device
+    }
+}}}
+
+
+#define OCV_INSTANTIATE_BILATERAL_FILTER(T) \
+    template void cv::gpu::device::imgproc::bilateral_filter_gpu<T>(const PtrStepSzb&, PtrStepSzb, int, float, float, int, cudaStream_t);
+
+OCV_INSTANTIATE_BILATERAL_FILTER(uchar)
+//OCV_INSTANTIATE_BILATERAL_FILTER(uchar2)
+OCV_INSTANTIATE_BILATERAL_FILTER(uchar3)
+OCV_INSTANTIATE_BILATERAL_FILTER(uchar4)
+
+//OCV_INSTANTIATE_BILATERAL_FILTER(schar)
+//OCV_INSTANTIATE_BILATERAL_FILTER(schar2)
+//OCV_INSTANTIATE_BILATERAL_FILTER(schar3)
+//OCV_INSTANTIATE_BILATERAL_FILTER(schar4)
+
+OCV_INSTANTIATE_BILATERAL_FILTER(short)
+//OCV_INSTANTIATE_BILATERAL_FILTER(short2)
+OCV_INSTANTIATE_BILATERAL_FILTER(short3)
+OCV_INSTANTIATE_BILATERAL_FILTER(short4)
+
+OCV_INSTANTIATE_BILATERAL_FILTER(ushort)
+//OCV_INSTANTIATE_BILATERAL_FILTER(ushort2)
+OCV_INSTANTIATE_BILATERAL_FILTER(ushort3)
+OCV_INSTANTIATE_BILATERAL_FILTER(ushort4)
+
+//OCV_INSTANTIATE_BILATERAL_FILTER(int)
+//OCV_INSTANTIATE_BILATERAL_FILTER(int2)
+//OCV_INSTANTIATE_BILATERAL_FILTER(int3)
+//OCV_INSTANTIATE_BILATERAL_FILTER(int4)
+
+OCV_INSTANTIATE_BILATERAL_FILTER(float)
+//OCV_INSTANTIATE_BILATERAL_FILTER(float2)
+OCV_INSTANTIATE_BILATERAL_FILTER(float3)
+OCV_INSTANTIATE_BILATERAL_FILTER(float4)
diff --git a/modules/gpu/src/cuda/nlm.cu b/modules/gpu/src/cuda/nlm.cu
new file mode 100644
index 0000000000..1acbe7fe7b
--- /dev/null
+++ b/modules/gpu/src/cuda/nlm.cu
@@ -0,0 +1,143 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 1993-2011, NVIDIA Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or bpied warranties, including, but not limited to, the bpied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "internal_shared.hpp"
+
+#include "opencv2/gpu/device/vec_traits.hpp"
+#include "opencv2/gpu/device/vec_math.hpp"
+#include "opencv2/gpu/device/border_interpolate.hpp"
+
+using namespace cv::gpu;
+
+typedef unsigned char uchar;
+typedef unsigned short ushort;
+
+//////////////////////////////////////////////////////////////////////////////////
+/// Non local means denosings
+
+namespace cv { namespace gpu { namespace device
+{
+    namespace imgproc
+    {
+        __device__ __forceinline__ float norm2(const float& v) { return v*v; }
+        __device__ __forceinline__ float norm2(const float2& v) { return v.x*v.x + v.y*v.y; }
+        __device__ __forceinline__ float norm2(const float3& v) { return v.x*v.x + v.y*v.y + v.z*v.z; }
+        __device__ __forceinline__ float norm2(const float4& v) { return v.x*v.x + v.y*v.y + v.z*v.z  + v.w*v.w; }
+
+        template<typename T, typename B>
+        __global__ void nlm_kernel(const PtrStepSz<T> src, PtrStep<T> dst, const B b, int search_radius, int block_radius, float h2_inv_half)
+        {
+            typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type value_type;
+
+            const int x = blockDim.x * blockIdx.x + threadIdx.x;
+            const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+            if (x >= src.cols || y >= src.rows)
+                return;
+
+            float block_radius2_inv = -1.f/(block_radius * block_radius);
+
+            value_type sum1 = VecTraits<value_type>::all(0);
+            float sum2 = 0.f;
+
+            for(float cy = -search_radius; cy <= search_radius; ++cy)
+                for(float cx = -search_radius; cx <= search_radius; ++cx)
+                {
+                    float color2 = 0;
+                    for(float by = -block_radius; by <= block_radius; ++by)
+                        for(float bx = -block_radius; bx <= block_radius; ++bx)
+                        {
+                            value_type v1 = saturate_cast<value_type>(src(y + by, x + bx));
+                            value_type v2 = saturate_cast<value_type>(src(y + cy + by, x + cx + bx));
+                            color2 += norm2(v1 - v2);
+                        }
+
+                    float dist2 = cx * cx + cy * cy;
+                    float w = __expf(color2 * h2_inv_half + dist2 * block_radius2_inv);
+                    
+                    sum1 = sum1 + saturate_cast<value_type>(src(y + cy, x + cy)) * w;
+                    sum2 += w;
+                }
+
+            dst(y, x) = saturate_cast<T>(sum1 / sum2);
+
+        }
+
+        template<typename T, template <typename> class B>
+        void nlm_caller(const PtrStepSzb src, PtrStepSzb dst, int search_radius, int block_radius, float h, cudaStream_t stream)
+        {
+            dim3 block (32, 8);
+            dim3 grid (divUp (src.cols, block.x), divUp (src.rows, block.y));
+
+            B<T> b(src.rows, src.cols);
+
+            float h2_inv_half = -0.5f/(h * h * VecTraits<T>::cn);
+
+            cudaSafeCall( cudaFuncSetCacheConfig (nlm_kernel<T, B<T> >, cudaFuncCachePreferL1) );
+            nlm_kernel<<<grid, block>>>((PtrStepSz<T>)src, (PtrStepSz<T>)dst, b, search_radius, block_radius, h2_inv_half);
+            cudaSafeCall ( cudaGetLastError () );
+
+            if (stream == 0)
+                cudaSafeCall( cudaDeviceSynchronize() );
+        }
+
+        template<typename T>
+        void nlm_bruteforce_gpu(const PtrStepSzb& src, PtrStepSzb dst, int search_radius, int block_radius, float h, int borderMode, cudaStream_t stream)
+        {
+            typedef void (*func_t)(const PtrStepSzb src, PtrStepSzb dst, int search_radius, int block_radius, float h, cudaStream_t stream);
+
+            static func_t funcs[] = 
+            {
+                nlm_caller<T, BrdReflect101>,
+                nlm_caller<T, BrdReplicate>,
+                nlm_caller<T, BrdConstant>,
+                nlm_caller<T, BrdReflect>,
+                nlm_caller<T, BrdWrap>,
+            };
+            funcs[borderMode](src, dst, search_radius, block_radius, h, stream);
+        }
+
+        template void nlm_bruteforce_gpu<uchar>(const PtrStepSzb&, PtrStepSzb, int, int, float, int, cudaStream_t);
+        template void nlm_bruteforce_gpu<uchar3>(const PtrStepSzb&, PtrStepSzb, int, int, float, int, cudaStream_t);
+    }
+}}}
diff --git a/modules/gpu/src/denoising.cpp b/modules/gpu/src/denoising.cpp
new file mode 100644
index 0000000000..f7dd2fbfa9
--- /dev/null
+++ b/modules/gpu/src/denoising.cpp
@@ -0,0 +1,135 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other GpuMaterials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or bpied warranties, including, but not limited to, the bpied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+using namespace cv;
+using namespace cv::gpu;
+
+#if !defined (HAVE_CUDA)
+
+cv::gpu::bilateralFilter(const GpuMat&, GpuMat&, int, float, float, int, Stream&) { throw_nogpu(); }
+
+#else
+
+
+namespace cv { namespace gpu { namespace device
+{
+    namespace imgproc
+    {
+        template<typename T>
+        void bilateral_filter_gpu(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t stream);
+
+        template<typename T>
+        void nlm_bruteforce_gpu(const PtrStepSzb& src, PtrStepSzb dst, int search_radius, int block_radius, float h, int borderMode, cudaStream_t stream);
+    }
+}}}
+
+void cv::gpu::bilateralFilter(const GpuMat& src, GpuMat& dst, int kernel_size, float sigma_color, float sigma_spatial, int borderMode, Stream& s)
+{
+    using cv::gpu::device::imgproc::bilateral_filter_gpu;
+
+    typedef void (*func_t)(const PtrStepSzb& src, PtrStepSzb dst, int kernel_size, float sigma_spatial, float sigma_color, int borderMode, cudaStream_t s);
+
+    static const func_t funcs[6][4] =
+    {
+        {bilateral_filter_gpu<uchar>      , 0 /*bilateral_filter_gpu<uchar2>*/ , bilateral_filter_gpu<uchar3>      , bilateral_filter_gpu<uchar4>      },
+        {0 /*bilateral_filter_gpu<schar>*/, 0 /*bilateral_filter_gpu<schar2>*/ , 0 /*bilateral_filter_gpu<schar3>*/, 0 /*bilateral_filter_gpu<schar4>*/},
+        {bilateral_filter_gpu<ushort>     , 0 /*bilateral_filter_gpu<ushort2>*/, bilateral_filter_gpu<ushort3>     , bilateral_filter_gpu<ushort4>     },
+        {bilateral_filter_gpu<short>      , 0 /*bilateral_filter_gpu<short2>*/ , bilateral_filter_gpu<short3>      , bilateral_filter_gpu<short4>      },
+        {0 /*bilateral_filter_gpu<int>*/  , 0 /*bilateral_filter_gpu<int2>*/   , 0 /*bilateral_filter_gpu<int3>*/  , 0 /*bilateral_filter_gpu<int4>*/  },
+        {bilateral_filter_gpu<float>      , 0 /*bilateral_filter_gpu<float2>*/ , bilateral_filter_gpu<float3>      , bilateral_filter_gpu<float4>      }
+    };
+
+    sigma_color = (sigma_color <= 0 ) ? 1 : sigma_color;
+    sigma_spatial = (sigma_spatial <= 0 ) ? 1 : sigma_spatial;
+    
+    
+    int radius = (kernel_size <= 0) ? cvRound(sigma_spatial*1.5) : kernel_size/2;
+    kernel_size = std::max(radius, 1)*2 + 1;
+
+    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    const func_t func = funcs[src.depth()][src.channels() - 1];
+    CV_Assert(func != 0);
+
+    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
+
+    int gpuBorderType;
+    CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
+
+    dst.create(src.size(), src.type());
+    func(src, dst, kernel_size, sigma_spatial, sigma_color, gpuBorderType, StreamAccessor::getStream(s));
+}
+
+void cv::gpu::nonLocalMeans(const GpuMat& src, GpuMat& dst, float h, int search_window_size, int block_size, int borderMode, Stream& s)
+{
+    using cv::gpu::device::imgproc::nlm_bruteforce_gpu;
+    typedef void (*func_t)(const PtrStepSzb& src, PtrStepSzb dst, int search_radius, int block_radius, float h, int borderMode, cudaStream_t stream);
+
+    static const func_t funcs[4] = { nlm_bruteforce_gpu<uchar>, 0 /*nlm_bruteforce_gpu<uchar2>*/ , nlm_bruteforce_gpu<uchar3>, 0/*nlm_bruteforce_gpu<uchar4>,*/ };
+
+    CV_Assert(src.type() == CV_8U || src.type() == CV_8UC3);
+
+    const func_t func = funcs[src.channels() - 1];
+    CV_Assert(func != 0);
+
+    int b = borderMode;
+    CV_Assert(b == BORDER_REFLECT101 || b == BORDER_REPLICATE || b == BORDER_CONSTANT || b == BORDER_REFLECT || b == BORDER_WRAP);
+
+    int gpuBorderType;
+    CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
+
+    int search_radius = search_window_size/2;
+    int block_radius = block_size/2;
+
+    dst.create(src.size(), src.type());
+    func(src, dst, search_radius, block_radius, h, gpuBorderType, StreamAccessor::getStream(s));
+}
+
+
+
+
+
+
+
+
+#endif
\ No newline at end of file
diff --git a/modules/gpu/src/hough.cpp b/modules/gpu/src/hough.cpp
index fd53057006..399de3684f 100644
--- a/modules/gpu/src/hough.cpp
+++ b/modules/gpu/src/hough.cpp
@@ -239,8 +239,8 @@ void cv::gpu::HoughCircles(const GpuMat& src, GpuMat& circles, HoughCirclesBuf&
 
                     for(size_t j = 0; j < m.size(); ++j)
                     {
-                        float dx = p.x - m[j].x;
-                        float dy = p.y - m[j].y;
+                        float dx = (float)(p.x - m[j].x);
+                        float dy = (float)(p.y - m[j].y);
 
                         if (dx * dx + dy * dy < minDist)
                         {
diff --git a/modules/gpu/src/opencv2/gpu/device/functional.hpp b/modules/gpu/src/opencv2/gpu/device/functional.hpp
index 1b836c7a56..96e96bded7 100644
--- a/modules/gpu/src/opencv2/gpu/device/functional.hpp
+++ b/modules/gpu/src/opencv2/gpu/device/functional.hpp
@@ -47,6 +47,7 @@
 #include "saturate_cast.hpp"
 #include "vec_traits.hpp"
 #include "type_traits.hpp"
+#include "device_functions.h"
 
 namespace cv { namespace gpu { namespace device
 {
@@ -408,6 +409,7 @@ namespace cv { namespace gpu { namespace device
     OPENCV_GPU_IMPLEMENT_BIN_FUNCTOR(pow, ::pow)
 
     #undef OPENCV_GPU_IMPLEMENT_UN_FUNCTOR
+    #undef OPENCV_GPU_IMPLEMENT_UN_FUNCTOR_NO_DOUBLE
     #undef OPENCV_GPU_IMPLEMENT_BIN_FUNCTOR
 
     template<typename T> struct hypot_sqr_func : binary_function<T, T, float>
diff --git a/modules/gpu/test/test_denoising.cpp b/modules/gpu/test/test_denoising.cpp
new file mode 100644
index 0000000000..0f6cd69c57
--- /dev/null
+++ b/modules/gpu/test/test_denoising.cpp
@@ -0,0 +1,140 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                        Intel License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000, Intel Corporation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "test_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+////////////////////////////////////////////////////////
+// BilateralFilter
+
+PARAM_TEST_CASE(BilateralFilter, cv::gpu::DeviceInfo, cv::Size, MatType)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int kernel_size;
+    float sigma_color;
+    float sigma_spatial;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+
+        kernel_size = 5;
+        sigma_color = 10.f;
+        sigma_spatial = 3.5f;
+
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+TEST_P(BilateralFilter, Accuracy)
+{
+    cv::Mat src = randomMat(size, type);
+    //cv::Mat src = readImage("hog/road.png", cv::IMREAD_GRAYSCALE);
+    //cv::Mat src = readImage("csstereobp/aloe-R.png", cv::IMREAD_GRAYSCALE);
+
+    src.convertTo(src, type);
+    cv::gpu::GpuMat dst;
+
+    cv::gpu::bilateralFilter(loadMat(src), dst, kernel_size, sigma_color, sigma_spatial);
+
+    cv::Mat dst_gold;
+    cv::bilateralFilter(src, dst_gold, kernel_size, sigma_color, sigma_spatial);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, src.depth() == CV_32F ? 1e-3 : 1.0);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, BilateralFilter, testing::Combine(
+    ALL_DEVICES,
+    testing::Values(cv::Size(128, 128), cv::Size(113, 113), cv::Size(639, 481)),
+    testing::Values(MatType(CV_8UC1), MatType(CV_8UC3), MatType(CV_32FC1), MatType(CV_32FC3))
+    ));
+
+
+////////////////////////////////////////////////////////
+// Brute Force Non local means
+
+struct NonLocalMeans: testing::TestWithParam<cv::gpu::DeviceInfo>
+{
+    cv::gpu::DeviceInfo devInfo;
+
+    virtual void SetUp()
+    {
+        devInfo = GetParam();
+        cv::gpu::setDevice(devInfo.deviceID());
+    }
+};
+
+TEST_P(NonLocalMeans, Regression)
+{
+    using cv::gpu::GpuMat;
+
+    cv::Mat bgr  = readImage("denoising/lena_noised_gaussian_sigma=20_multi_0.png", cv::IMREAD_COLOR);
+    ASSERT_FALSE(bgr.empty());
+    
+    cv::Mat gray;
+    cv::cvtColor(bgr, gray, CV_BGR2GRAY);
+
+    GpuMat dbgr, dgray;
+    cv::gpu::nonLocalMeans(GpuMat(bgr),  dbgr, 10);
+    cv::gpu::nonLocalMeans(GpuMat(gray), dgray, 10);
+
+#if 0
+    dumpImage("denoising/denoised_lena_bgr.png", cv::Mat(dbgr));
+    dumpImage("denoising/denoised_lena_gray.png", cv::Mat(dgray));
+#endif
+
+    cv::Mat bgr_gold  = readImage("denoising/denoised_lena_bgr.png", cv::IMREAD_COLOR);
+    cv::Mat gray_gold  = readImage("denoising/denoised_lena_gray.png", cv::IMREAD_GRAYSCALE);
+    ASSERT_FALSE(bgr_gold.empty() || gray_gold.empty());
+
+    EXPECT_MAT_NEAR(bgr_gold, dbgr, 1e-4);
+    EXPECT_MAT_NEAR(gray_gold, dgray, 1e-4);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_ImgProc, NonLocalMeans, ALL_DEVICES);
+
+
+#endif // HAVE_CUDA
\ No newline at end of file
diff --git a/modules/gpu/test/utility.cpp b/modules/gpu/test/utility.cpp
index cf3b0fc8cb..a92d2c52f1 100644
--- a/modules/gpu/test/utility.cpp
+++ b/modules/gpu/test/utility.cpp
@@ -127,6 +127,14 @@ Mat readImageType(const std::string& fname, int type)
     return src;
 }
 
+//////////////////////////////////////////////////////////////////////
+// Image dumping
+
+void dumpImage(const std::string& fileName, const cv::Mat& image)
+{
+    cv::imwrite(TS::ptr()->get_data_path() + fileName, image);
+}
+
 //////////////////////////////////////////////////////////////////////
 // Gpu devices
 
diff --git a/modules/gpu/test/utility.hpp b/modules/gpu/test/utility.hpp
index f509b786ad..1d153fcf02 100644
--- a/modules/gpu/test/utility.hpp
+++ b/modules/gpu/test/utility.hpp
@@ -74,6 +74,11 @@ cv::Mat readImage(const std::string& fileName, int flags = cv::IMREAD_COLOR);
 //! read image from testdata folder and convert it to specified type
 cv::Mat readImageType(const std::string& fname, int type);
 
+//////////////////////////////////////////////////////////////////////
+// Image dumping
+
+void dumpImage(const std::string& fileName, const cv::Mat& image);
+
 //////////////////////////////////////////////////////////////////////
 // Gpu devices
 
diff --git a/modules/imgproc/src/smooth.cpp b/modules/imgproc/src/smooth.cpp
index 8ec6f6d793..13340511f9 100644
--- a/modules/imgproc/src/smooth.cpp
+++ b/modules/imgproc/src/smooth.cpp
@@ -1285,6 +1285,8 @@ void cv::medianBlur( InputArray _src0, OutputArray _dst, int ksize )
                                    Bilateral Filtering
 \****************************************************************************************/
 
+#undef CV_SSE3
+
 namespace cv
 {