added buffered version of cornerHarris, cornerMinEigenVal, histEven and histRange

diff --git a/modules/gpu/include/opencv2/gpu/gpu.hpp b/modules/gpu/include/opencv2/gpu/gpu.hpp
index c733066..168433d 100644 (file)
--- a/modules/gpu/include/opencv2/gpu/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu/gpu.hpp
@@ -862,9 +862,11 @@ namespace cv
 \r
         //! computes Harris cornerness criteria at each image pixel\r
         CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType=BORDER_REFLECT101);\r
+        CV_EXPORTS void cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType=BORDER_REFLECT101);\r
 \r
         //! computes minimum eigen value of 2x2 derivative covariation matrix at each pixel - the cornerness criteria\r
         CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType=BORDER_REFLECT101);\r
+        CV_EXPORTS void cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType=BORDER_REFLECT101);\r
 \r
         //! performs per-element multiplication of two full (not packed) Fourier spectrums\r
         //! supports 32FC2 matrixes only (interleaved format)\r
@@ -1096,22 +1098,26 @@ namespace cv
         //! Supports CV_8UC1, CV_16UC1 and CV_16SC1 source types.\r
         //! Output hist will have one row and histSize cols and CV_32SC1 type.\r
         CV_EXPORTS void histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());\r
+        CV_EXPORTS void histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream = Stream::Null());\r
         //! Calculates histogram with evenly distributed bins for four-channel source.\r
         //! All channels of source are processed separately.\r
         //! Supports CV_8UC4, CV_16UC4 and CV_16SC4 source types.\r
         //! Output hist[i] will have one row and histSize[i] cols and CV_32SC1 type.\r
         CV_EXPORTS void histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());\r
+        CV_EXPORTS void histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream = Stream::Null());\r
         //! Calculates histogram with bins determined by levels array.\r
         //! levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.\r
         //! Supports CV_8UC1, CV_16UC1, CV_16SC1 and CV_32FC1 source types.\r
         //! Output hist will have one row and (levels.cols-1) cols and CV_32SC1 type.\r
         CV_EXPORTS void histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream = Stream::Null());\r
+        CV_EXPORTS void histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream = Stream::Null());\r
         //! Calculates histogram with bins determined by levels array.\r
         //! All levels must have one row and CV_32SC1 type if source has integer type or CV_32FC1 otherwise.\r
         //! All channels of source are processed separately.\r
         //! Supports CV_8UC4, CV_16UC4, CV_16SC4 and CV_32FC4 source types.\r
         //! Output hist[i] will have one row and (levels[i].cols-1) cols and CV_32SC1 type.\r
         CV_EXPORTS void histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream = Stream::Null());\r
+        CV_EXPORTS void histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream = Stream::Null());\r
         \r
         //! Calculates histogram for 8u one channel image\r
         //! Output hist will have one row, 256 cols and CV32SC1 type.\r

diff --git a/modules/gpu/src/imgproc_gpu.cpp b/modules/gpu/src/imgproc.cpp
similarity index 92%
rename from modules/gpu/src/imgproc_gpu.cpp
rename to modules/gpu/src/imgproc.cpp
index 8318513..de35c39 100644 (file)
--- a/modules/gpu/src/imgproc_gpu.cpp
+++ b/modules/gpu/src/imgproc.cpp
@@ -68,16 +68,22 @@ void cv::gpu::columnSum(const GpuMat&, GpuMat&) { throw_nogpu(); }
 void cv::gpu::rectStdDev(const GpuMat&, const GpuMat&, GpuMat&, const Rect&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::evenLevels(GpuMat&, int, int, int) { throw_nogpu(); }\r
 void cv::gpu::histEven(const GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }\r
+void cv::gpu::histEven(const GpuMat&, GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }\r
 void cv::gpu::histEven(const GpuMat&, GpuMat*, int*, int*, int*, Stream&) { throw_nogpu(); }\r
+void cv::gpu::histEven(const GpuMat&, GpuMat*, GpuMat&, int*, int*, int*, Stream&) { throw_nogpu(); }\r
 void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }\r
+void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, Stream&) { throw_nogpu(); }\r
+void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::calcHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::calcHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }\r
 void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }\r
+void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }\r
 void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }\r
+void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }\r
 void cv::gpu::mulSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, bool) { throw_nogpu(); }\r
 void cv::gpu::mulAndScaleSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, float, bool) { throw_nogpu(); }\r
 void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int) { throw_nogpu(); }\r
@@ -127,15 +133,9 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
 \r
     CV_Assert((src.type() == CV_8U || src.type() == CV_8UC3) && xmap.type() == CV_32F && ymap.type() == CV_32F);\r
 \r
-    GpuMat out;\r
-    if (dst.data != src.data)\r
-        out = dst;\r
+    dst.create(xmap.size(), src.type());\r
 \r
-    out.create(xmap.size(), src.type());\r
-\r
-    callers[src.channels() - 1](src, xmap, ymap, out);\r
-\r
-    dst = out;\r
+    callers[src.channels() - 1](src, xmap, ymap, dst);\r
 }\r
 \r
 ////////////////////////////////////////////////////////////////////////\r
@@ -199,14 +199,9 @@ namespace
     template <typename T>\r
     void drawColorDisp_caller(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream)\r
     {\r
-        GpuMat out;\r
-        if (dst.data != src.data)\r
-            out = dst;\r
-        out.create(src.size(), CV_8UC4);\r
-\r
-        imgproc::drawColorDisp_gpu((DevMem2D_<T>)src, out, ndisp, stream);\r
+        dst.create(src.size(), CV_8UC4);\r
 \r
-        dst = out;\r
+        imgproc::drawColorDisp_gpu((DevMem2D_<T>)src, dst, ndisp, stream);\r
     }\r
 \r
     typedef void (*drawColorDisp_caller_t)(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream);\r
@@ -806,7 +801,7 @@ namespace
     {\r
         typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;\r
 \r
-        static void hist(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)\r
+        static void hist(const GpuMat& src, GpuMat& hist, GpuMat& buffer, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)\r
         {\r
             int levels = histSize + 1;\r
             hist.create(1, histSize, CV_32S);\r
@@ -815,11 +810,10 @@ namespace
             sz.width = src.cols;\r
             sz.height = src.rows;\r
 \r
-            GpuMat buffer;\r
             int buf_size;\r
-\r
             get_buf_size(sz, levels, &buf_size);\r
-            buffer.create(1, buf_size, CV_8U);\r
+\r
+            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);\r
 \r
             NppStreamHandler h(stream);\r
 \r
@@ -835,7 +829,7 @@ namespace
     {\r
         typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;\r
 \r
-        static void hist(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)\r
+        static void hist(const GpuMat& src, GpuMat hist[4], GpuMat& buffer, int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)\r
         {\r
             int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};\r
             hist[0].create(1, histSize[0], CV_32S);\r
@@ -849,11 +843,10 @@ namespace
 \r
             Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};\r
 \r
-            GpuMat buffer;\r
             int buf_size;\r
-\r
             get_buf_size(sz, levels, &buf_size);\r
-            buffer.create(1, buf_size, CV_8U);\r
+\r
+            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);\r
 \r
             NppStreamHandler h(stream);\r
 \r
@@ -908,7 +901,7 @@ namespace
         typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;\r
         enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};\r
 \r
-        static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels, cudaStream_t stream)\r
+        static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buffer, cudaStream_t stream)\r
         {\r
             CV_Assert(levels.type() == LEVEL_TYPE_CODE && levels.rows == 1);\r
 \r
@@ -918,11 +911,10 @@ namespace
             sz.width = src.cols;\r
             sz.height = src.rows;\r
 \r
-            GpuMat buffer;\r
             int buf_size;\r
-\r
             get_buf_size(sz, levels.cols, &buf_size);\r
-            buffer.create(1, buf_size, CV_8U);\r
+            \r
+            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);\r
 \r
             NppStreamHandler h(stream);\r
 \r
@@ -939,7 +931,7 @@ namespace
         typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;\r
         enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};\r
 \r
-        static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], cudaStream_t stream)\r
+        static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buffer, cudaStream_t stream)\r
         {\r
             CV_Assert(levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1);\r
             CV_Assert(levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1);\r
@@ -959,11 +951,10 @@ namespace
             sz.width = src.cols;\r
             sz.height = src.rows;\r
 \r
-            GpuMat buffer;\r
             int buf_size;\r
-\r
             get_buf_size(sz, nLevels, &buf_size);\r
-            buffer.create(1, buf_size, CV_8U);\r
+\r
+            ensureSizeIsEnough(1, buf_size, CV_8U, buffer);\r
 \r
             NppStreamHandler h(stream);\r
 \r
@@ -984,9 +975,15 @@ void cv::gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperL
 \r
 void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream)\r
 {\r
+    GpuMat buf;\r
+    histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);\r
+}\r
+\r
+void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream)\r
+{\r
     CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );\r
 \r
-    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);\r
+    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, GpuMat& buf, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);\r
     static const hist_t hist_callers[] =\r
     {\r
         NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,\r
@@ -995,14 +992,20 @@ void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerL
         NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist\r
     };\r
 \r
-    hist_callers[src.depth()](src, hist, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));\r
+    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));\r
 }\r
 \r
 void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)\r
 {\r
+    GpuMat buf;\r
+    histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);\r
+}\r
+\r
+void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)\r
+{\r
     CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );\r
 \r
-    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], int levels[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream);\r
+    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int levels[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream);\r
     static const hist_t hist_callers[] =\r
     {\r
         NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,\r
@@ -1011,14 +1014,21 @@ void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int l
         NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist\r
     };\r
 \r
-    hist_callers[src.depth()](src, hist, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));\r
+    hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));\r
 }\r
 \r
 void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream)\r
 {\r
+    GpuMat buf;\r
+    histRange(src, hist, levels, buf, stream);\r
+}\r
+\r
+\r
+void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream)\r
+{\r
     CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1);\r
 \r
-    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels, cudaStream_t stream);\r
+    typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, cudaStream_t stream);\r
     static const hist_t hist_callers[] =\r
     {\r
         NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,\r
@@ -1029,14 +1039,20 @@ void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, S
         NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist\r
     };\r
 \r
-    hist_callers[src.depth()](src, hist, levels, StreamAccessor::getStream(stream));\r
+    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));\r
 }\r
 \r
 void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream)\r
 {\r
+    GpuMat buf;\r
+    histRange(src, hist, levels, buf, stream);\r
+}\r
+\r
+void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream)\r
+{\r
     CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4);\r
 \r
-    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], cudaStream_t stream);\r
+    typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, cudaStream_t stream);\r
     static const hist_t hist_callers[] =\r
     {\r
         NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,\r
@@ -1047,7 +1063,7 @@ void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4
         NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist\r
     };\r
 \r
-    hist_callers[src.depth()](src, hist, levels, StreamAccessor::getStream(stream));\r
+    hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));\r
 }\r
 \r
 namespace cv { namespace gpu { namespace histograms\r
@@ -1151,7 +1167,6 @@ namespace
             scale *= 255.;\r
         scale = 1./scale;\r
 \r
-        GpuMat tmp_buf(src.size(), CV_32F);\r
         Dx.create(src.size(), CV_32F);\r
         Dy.create(src.size(), CV_32F);\r
 \r
@@ -1210,13 +1225,18 @@ bool cv::gpu::tryConvertToGpuBorderType(int cpuBorderType, int& gpuBorderType)
 \r
 void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType)\r
 {\r
+    GpuMat Dx, Dy;\r
+    cornerHarris(src, dst, Dx, Dy, blockSize, ksize, k, borderType);\r
+}\r
+\r
+void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType)\r
+{\r
     CV_Assert(borderType == cv::BORDER_REFLECT101 ||\r
               borderType == cv::BORDER_REPLICATE);\r
 \r
     int gpuBorderType;\r
     CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));\r
 \r
-    GpuMat Dx, Dy;\r
     extractCovData(src, Dx, Dy, blockSize, ksize, borderType);\r
     dst.create(src.size(), CV_32F);\r
     imgproc::cornerHarris_caller(blockSize, (float)k, Dx, Dy, dst, gpuBorderType);\r
@@ -1224,13 +1244,18 @@ void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ks
 \r
 void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType)\r
 {  \r
+    GpuMat Dx, Dy;\r
+    cornerMinEigenVal(src, dst, Dx, Dy, blockSize, ksize, borderType);\r
+}\r
+\r
+void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)\r
+{  \r
     CV_Assert(borderType == cv::BORDER_REFLECT101 ||\r
               borderType == cv::BORDER_REPLICATE);\r
 \r
     int gpuBorderType;\r
     CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));\r
 \r
-    GpuMat Dx, Dy;\r
     extractCovData(src, Dx, Dy, blockSize, ksize, borderType);    \r
     dst.create(src.size(), CV_32F);\r
     imgproc::cornerMinEigenVal_caller(blockSize, Dx, Dy, dst, gpuBorderType);\r