Merge pull request #979 from jet47:gpuwarping-refactoring

modules/gpuwarping/doc/warping.rst

@@ -9,7 +9,7 @@ gpu::remap
 --------------
 Applies a generic geometrical transformation to an image.
 
-.. ocv:function:: void gpu::remap( const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
+.. ocv:function:: void gpu::remap(InputArray src, OutputArray dst, InputArray xmap, InputArray ymap, int interpolation, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
 
     :param src: Source image.
 
@@ -43,7 +43,7 @@ gpu::resize
 ---------------
 Resizes an image.
 
-.. ocv:function:: void gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
 
     :param src: Source image.
 
@@ -80,7 +80,7 @@ gpu::warpAffine
 -------------------
 Applies an affine transformation to an image.
 
-.. ocv:function:: void gpu::warpAffine( const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags=INTER_LINEAR, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
+.. ocv:function:: void gpu::warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
 
     :param src: Source image.  ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
 
@@ -102,7 +102,7 @@ gpu::buildWarpAffineMaps
 ------------------------
 Builds transformation maps for affine transformation.
 
-.. ocv:function:: void gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null())
 
     :param M: *2x3*  transformation matrix.
 
@@ -124,7 +124,7 @@ gpu::warpPerspective
 ------------------------
 Applies a perspective transformation to an image.
 
-.. ocv:function:: void gpu::warpPerspective( const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags=INTER_LINEAR, int borderMode=BORDER_CONSTANT, Scalar borderValue=Scalar(), Stream& stream=Stream::Null() )
+.. ocv:function:: void gpu::warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null())
 
     :param src: Source image. ``CV_8U`` , ``CV_16U`` , ``CV_32S`` , or  ``CV_32F`` depth and 1, 3, or 4 channels are supported.
 
@@ -146,7 +146,7 @@ gpu::buildWarpPerspectiveMaps
 -----------------------------
 Builds transformation maps for perspective transformation.
 
-.. ocv:function:: void gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null())
 
     :param M: *3x3*  transformation matrix.
 
@@ -164,11 +164,41 @@ Builds transformation maps for perspective transformation.
 
 
 
+gpu::buildWarpPlaneMaps
+-----------------------
+Builds plane warping maps.
+
+.. ocv:function:: void gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, InputArray T, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::buildWarpCylindricalMaps
+-----------------------------
+Builds cylindrical warping maps.
+
+.. ocv:function:: void gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
+gpu::buildWarpSphericalMaps
+---------------------------
+Builds spherical warping maps.
+
+.. ocv:function:: void gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale, OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null())
+
+    :param stream: Stream for the asynchronous version.
+
+
+
 gpu::rotate
 ---------------
 Rotates an image around the origin (0,0) and then shifts it.
 
-.. ocv:function:: void gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift = 0, double yShift = 0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::rotate(InputArray src, OutputArray dst, Size dsize, double angle, double xShift = 0, double yShift = 0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null())
 
     :param src: Source image. Supports 1, 3 or 4 channels images with ``CV_8U`` , ``CV_16U`` or ``CV_32F`` depth.
 
@@ -190,41 +220,11 @@ Rotates an image around the origin (0,0) and then shifts it.
 
 
 
-gpu::buildWarpPlaneMaps
------------------------
-Builds plane warping maps.
-
-.. ocv:function:: void gpu::buildWarpPlaneMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, const Mat & T, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::buildWarpCylindricalMaps
------------------------------
-Builds cylindrical warping maps.
-
-.. ocv:function:: void gpu::buildWarpCylindricalMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
-gpu::buildWarpSphericalMaps
----------------------------
-Builds spherical warping maps.
-
-.. ocv:function:: void gpu::buildWarpSphericalMaps( Size src_size, Rect dst_roi, const Mat & K, const Mat& R, float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream=Stream::Null() )
-
-    :param stream: Stream for the asynchronous version.
-
-
-
 gpu::pyrDown
 -------------------
 Smoothes an image and downsamples it.
 
-.. ocv:function:: void gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::pyrDown(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
 
     :param src: Source image.
 
@@ -240,7 +240,7 @@ gpu::pyrUp
 -------------------
 Upsamples an image and then smoothes it.
 
-.. ocv:function:: void gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null())
+.. ocv:function:: void gpu::pyrUp(InputArray src, OutputArray dst, Stream& stream = Stream::Null())
 
     :param src: Source image.
 

modules/gpuwarping/include/opencv2/gpuwarping.hpp

@@ -54,78 +54,60 @@ namespace cv { namespace gpu {
 
 //! DST[x,y] = SRC[xmap[x,y],ymap[x,y]]
 //! supports only CV_32FC1 map type
-CV_EXPORTS void remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap,
+CV_EXPORTS void remap(InputArray src, OutputArray dst, InputArray xmap, InputArray ymap,
                       int interpolation, int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(),
                       Stream& stream = Stream::Null());
 
 //! resizes the image
 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_AREA
-CV_EXPORTS void resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
+CV_EXPORTS void resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
 
 //! warps the image using affine transformation
 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
-CV_EXPORTS void warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR,
+CV_EXPORTS void warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
     int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
 
-CV_EXPORTS void buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null());
+CV_EXPORTS void buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
 
 //! warps the image using perspective transformation
 //! Supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
-CV_EXPORTS void warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags = INTER_LINEAR,
+CV_EXPORTS void warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
     int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
 
-CV_EXPORTS void buildWarpPerspectiveMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream = Stream::Null());
+CV_EXPORTS void buildWarpPerspectiveMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
 
 //! builds plane warping maps
-CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, const Mat &T, float scale,
+CV_EXPORTS void buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, InputArray T, float scale,
-                                   GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
+                                   OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
 
 //! builds cylindrical warping maps
-CV_EXPORTS void buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
+CV_EXPORTS void buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale,
-                                         GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
+                                         OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
 
 //! builds spherical warping maps
-CV_EXPORTS void buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
+CV_EXPORTS void buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray K, InputArray R, float scale,
-                                       GpuMat& map_x, GpuMat& map_y, Stream& stream = Stream::Null());
+                                       OutputArray map_x, OutputArray map_y, Stream& stream = Stream::Null());
 
 //! rotates an image around the origin (0,0) and then shifts it
 //! supports INTER_NEAREST, INTER_LINEAR, INTER_CUBIC
 //! supports 1, 3 or 4 channels images with CV_8U, CV_16U or CV_32F depth
-CV_EXPORTS void rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift = 0, double yShift = 0,
+CV_EXPORTS void rotate(InputArray src, OutputArray dst, Size dsize, double angle, double xShift = 0, double yShift = 0,
                        int interpolation = INTER_LINEAR, Stream& stream = Stream::Null());
 
 //! smoothes the source image and downsamples it
-CV_EXPORTS void pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
+CV_EXPORTS void pyrDown(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
 
 //! upsamples the source image and then smoothes it
-CV_EXPORTS void pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream = Stream::Null());
+CV_EXPORTS void pyrUp(InputArray src, OutputArray dst, Stream& stream = Stream::Null());
 
-class CV_EXPORTS ImagePyramid
+class CV_EXPORTS ImagePyramid : public Algorithm
 {
 public:
-    inline ImagePyramid() : nLayers_(0) {}
+    virtual void getLayer(OutputArray outImg, Size outRoi, Stream& stream = Stream::Null()) const = 0;
-    inline ImagePyramid(const GpuMat& img, int nLayers, Stream& stream = Stream::Null())
-    {
-        build(img, nLayers, stream);
-    }
-
-    void build(const GpuMat& img, int nLayers, Stream& stream = Stream::Null());
-
-    void getLayer(GpuMat& outImg, Size outRoi, Stream& stream = Stream::Null()) const;
-
-    inline void release()
-    {
-        layer0_.release();
-        pyramid_.clear();
-        nLayers_ = 0;
-    }
-
-private:
-    GpuMat layer0_;
-    std::vector<GpuMat> pyramid_;
-    int nLayers_;
 };
 
+CV_EXPORTS Ptr<ImagePyramid> createImagePyramid(InputArray img, int nLayers = -1, Stream& stream = Stream::Null());
+
 }} // namespace cv { namespace gpu {
 
 #endif /* __OPENCV_GPUWARPING_HPP__ */

modules/gpuwarping/perf/perf_warping.cpp

@@ -515,45 +515,6 @@ PERF_TEST_P(Sz_Depth_Cn, PyrUp,
     }
 }
 
-//////////////////////////////////////////////////////////////////////
-// ImagePyramidBuild
-
-PERF_TEST_P(Sz_Depth_Cn, ImagePyramidBuild,
-            Combine(GPU_TYPICAL_MAT_SIZES,
-                    Values(CV_8U, CV_16U, CV_32F),
-                    GPU_CHANNELS_1_3_4))
-{
-    const cv::Size size = GET_PARAM(0);
-    const int depth = GET_PARAM(1);
-    const int channels = GET_PARAM(2);
-
-    const int type = CV_MAKE_TYPE(depth, channels);
-
-    cv::Mat src(size, type);
-    declare.in(src, WARMUP_RNG);
-
-    const int nLayers = 5;
-    const cv::Size dstSize(size.width / 2 + 10, size.height / 2 + 10);
-
-    if (PERF_RUN_GPU())
-    {
-        const cv::gpu::GpuMat d_src(src);
-
-        cv::gpu::ImagePyramid d_pyr;
-
-        TEST_CYCLE() d_pyr.build(d_src, nLayers);
-
-        cv::gpu::GpuMat dst;
-        d_pyr.getLayer(dst, dstSize);
-
-        GPU_SANITY_CHECK(dst);
-    }
-    else
-    {
-        FAIL_NO_CPU();
-    }
-}
-
 //////////////////////////////////////////////////////////////////////
 // ImagePyramidGetLayer
 
@@ -579,9 +540,9 @@ PERF_TEST_P(Sz_Depth_Cn, ImagePyramidGetLayer,
         const cv::gpu::GpuMat d_src(src);
         cv::gpu::GpuMat dst;
 
-        cv::gpu::ImagePyramid d_pyr(d_src, nLayers);
+        cv::Ptr<cv::gpu::ImagePyramid> d_pyr = cv::gpu::createImagePyramid(d_src, nLayers);
 
-        TEST_CYCLE() d_pyr.getLayer(dst, dstSize);
+        TEST_CYCLE() d_pyr->getLayer(dst, dstSize);
 
         GPU_SANITY_CHECK(dst);
     }

modules/gpuwarping/src/pyramids.cpp

@@ -42,14 +42,15 @@
 
 #include "precomp.hpp"
 
+using namespace cv;
+using namespace cv::gpu;
+
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::pyrDown(const GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::pyrDown(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::gpu::pyrUp(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::pyrUp(const GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+Ptr<ImagePyramid> cv::gpu::createImagePyramid(InputArray, int, Stream&) { throw_no_cuda(); return Ptr<ImagePyramid>(); }
-
-void cv::gpu::ImagePyramid::build(const GpuMat&, int, Stream&) { throw_no_cuda(); }
-void cv::gpu::ImagePyramid::getLayer(GpuMat&, Size, Stream&) const { throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
@@ -64,12 +65,11 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::pyrDown(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace cv::gpu::cudev::imgproc;
 
     typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
     static const func_t funcs[6][4] =
     {
         {pyrDown_gpu<uchar>      , 0 /*pyrDown_gpu<uchar2>*/ , pyrDown_gpu<uchar3>      , pyrDown_gpu<uchar4>      },
@@ -80,12 +80,15 @@ void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream)
         {pyrDown_gpu<float>      , 0 /*pyrDown_gpu<float2>*/ , pyrDown_gpu<float3>      , pyrDown_gpu<float4>      }
     };
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    GpuMat src = _src.getGpuMat();
+
+    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( func != 0 );
 
-    dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
+    _dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     func(src, dst, StreamAccessor::getStream(stream));
 }
@@ -102,12 +105,11 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream)
+void cv::gpu::pyrUp(InputArray _src, OutputArray _dst, Stream& stream)
 {
     using namespace cv::gpu::cudev::imgproc;
 
     typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
-
     static const func_t funcs[6][4] =
     {
         {pyrUp_gpu<uchar>      , 0 /*pyrUp_gpu<uchar2>*/ , pyrUp_gpu<uchar3>      , pyrUp_gpu<uchar4>      },
@@ -118,98 +120,124 @@ void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream)
         {pyrUp_gpu<float>      , 0 /*pyrUp_gpu<float2>*/ , pyrUp_gpu<float3>      , pyrUp_gpu<float4>      }
     };
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    GpuMat src = _src.getGpuMat();
+
+    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( func != 0 );
 
-    dst.create(src.rows * 2, src.cols * 2, src.type());
+    _dst.create(src.rows * 2, src.cols * 2, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     func(src, dst, StreamAccessor::getStream(stream));
 }
 
-
 //////////////////////////////////////////////////////////////////////////////
 // ImagePyramid
 
-void cv::gpu::ImagePyramid::build(const GpuMat& img, int numLayers, Stream& stream)
+#ifdef HAVE_OPENCV_GPULEGACY
+
+namespace
+{
+    class ImagePyramidImpl : public ImagePyramid
+    {
+    public:
+        ImagePyramidImpl(InputArray img, int nLayers, Stream& stream);
+
+        void getLayer(OutputArray outImg, Size outRoi, Stream& stream = Stream::Null()) const;
+
+    private:
+        GpuMat layer0_;
+        std::vector<GpuMat> pyramid_;
+        int nLayers_;
+    };
+
+    ImagePyramidImpl::ImagePyramidImpl(InputArray _img, int numLayers, Stream& stream)
+    {
+        GpuMat img = _img.getGpuMat();
+
+        CV_Assert( img.depth() <= CV_32F && img.channels() <= 4 );
+
+        img.copyTo(layer0_, stream);
+
+        Size szLastLayer = img.size();
+        nLayers_ = 1;
+
+        if (numLayers <= 0)
+            numLayers = 255; // it will cut-off when any of the dimensions goes 1
+
+        pyramid_.resize(numLayers);
+
+        for (int i = 0; i < numLayers - 1; ++i)
+        {
+            Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
+
+            if (szCurLayer.width == 0 || szCurLayer.height == 0)
+                break;
+
+            ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
+            nLayers_++;
+
+            const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
+
+            cudev::pyramid::downsampleX2(prevLayer, pyramid_[i], img.depth(), img.channels(), StreamAccessor::getStream(stream));
+
+            szLastLayer = szCurLayer;
+        }
+    }
+
+    void ImagePyramidImpl::getLayer(OutputArray _outImg, Size outRoi, Stream& stream) const
+    {
+        CV_Assert( outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0 );
+
+        ensureSizeIsEnough(outRoi, layer0_.type(), _outImg);
+        GpuMat outImg = _outImg.getGpuMat();
+
+        if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
+        {
+            layer0_.copyTo(outImg, stream);
+            return;
+        }
+
+        float lastScale = 1.0f;
+        float curScale;
+        GpuMat lastLayer = layer0_;
+        GpuMat curLayer;
+
+        for (int i = 0; i < nLayers_ - 1; ++i)
+        {
+            curScale = lastScale * 0.5f;
+            curLayer = pyramid_[i];
+
+            if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
+            {
+                curLayer.copyTo(outImg, stream);
+            }
+
+            if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
+                break;
+
+            lastScale = curScale;
+            lastLayer = curLayer;
+        }
+
+        cudev::pyramid::interpolateFrom1(lastLayer, outImg, outImg.depth(), outImg.channels(), StreamAccessor::getStream(stream));
+    }
+}
+
+#endif
+
+Ptr<ImagePyramid> cv::gpu::createImagePyramid(InputArray img, int nLayers, Stream& stream)
 {
 #ifndef HAVE_OPENCV_GPULEGACY
     (void) img;
     (void) numLayers;
     (void) stream;
     throw_no_cuda();
+    return Ptr<ImagePyramid>();
 #else
-    CV_Assert(img.depth() <= CV_32F && img.channels() <= 4);
+    return new ImagePyramidImpl(img, nLayers, stream);
-
-    layer0_ = img;
-    Size szLastLayer = img.size();
-    nLayers_ = 1;
-
-    if (numLayers <= 0)
-        numLayers = 255; //it will cut-off when any of the dimensions goes 1
-
-    pyramid_.resize(numLayers);
-
-    for (int i = 0; i < numLayers - 1; ++i)
-    {
-        Size szCurLayer(szLastLayer.width / 2, szLastLayer.height / 2);
-
-        if (szCurLayer.width == 0 || szCurLayer.height == 0)
-            break;
-
-        ensureSizeIsEnough(szCurLayer, img.type(), pyramid_[i]);
-        nLayers_++;
-
-        const GpuMat& prevLayer = i == 0 ? layer0_ : pyramid_[i - 1];
-
-        cudev::pyramid::downsampleX2(prevLayer, pyramid_[i], img.depth(), img.channels(), StreamAccessor::getStream(stream));
-
-        szLastLayer = szCurLayer;
-    }
-#endif
-}
-
-void cv::gpu::ImagePyramid::getLayer(GpuMat& outImg, Size outRoi, Stream& stream) const
-{
-#ifndef HAVE_OPENCV_GPULEGACY
-    (void) outImg;
-    (void) outRoi;
-    (void) stream;
-    throw_no_cuda();
-#else
-    CV_Assert(outRoi.width <= layer0_.cols && outRoi.height <= layer0_.rows && outRoi.width > 0 && outRoi.height > 0);
-
-    ensureSizeIsEnough(outRoi, layer0_.type(), outImg);
-
-    if (outRoi.width == layer0_.cols && outRoi.height == layer0_.rows)
-    {
-        layer0_.copyTo(outImg, stream);
-    }
-
-    float lastScale = 1.0f;
-    float curScale;
-    GpuMat lastLayer = layer0_;
-    GpuMat curLayer;
-
-    for (int i = 0; i < nLayers_ - 1; ++i)
-    {
-        curScale = lastScale * 0.5f;
-        curLayer = pyramid_[i];
-
-        if (outRoi.width == curLayer.cols && outRoi.height == curLayer.rows)
-        {
-            curLayer.copyTo(outImg, stream);
-        }
-
-        if (outRoi.width >= curLayer.cols && outRoi.height >= curLayer.rows)
-            break;
-
-        lastScale = curScale;
-        lastLayer = curLayer;
-    }
-
-    cudev::pyramid::interpolateFrom1(lastLayer, outImg, outImg.depth(), outImg.channels(), StreamAccessor::getStream(stream));
 #endif
 }
 

modules/gpuwarping/src/remap.cpp

@@ -44,7 +44,7 @@
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::remap(const GpuMat&, GpuMat&, const GpuMat&, const GpuMat&, int, int, Scalar, Stream&){ throw_no_cuda(); }
+void cv::gpu::remap(InputArray, OutputArray, InputArray, InputArray, int, int, Scalar, Stream&){ throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
@@ -58,13 +58,12 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode, Scalar borderValue, Stream& stream)
+void cv::gpu::remap(InputArray _src, OutputArray _dst, InputArray _xmap, InputArray _ymap, int interpolation, int borderMode, Scalar borderValue, Stream& stream)
 {
     using namespace cv::gpu::cudev::imgproc;
 
     typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation,
         int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
-
     static const func_t funcs[6][4] =
     {
         {remap_gpu<uchar>      , 0 /*remap_gpu<uchar2>*/ , remap_gpu<uchar3>     , remap_gpu<uchar4>     },
@@ -75,15 +74,21 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
         {remap_gpu<float>      , 0 /*remap_gpu<float2>*/ , remap_gpu<float3>     , remap_gpu<float4>     }
     };
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    GpuMat src = _src.getGpuMat();
-    CV_Assert(xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size());
+    GpuMat xmap = _xmap.getGpuMat();
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
+    GpuMat ymap = _ymap.getGpuMat();
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
+
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size() );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
 
     const func_t func = funcs[src.depth()][src.channels() - 1];
-    CV_Assert(func != 0);
+    if (!func)
+        CV_Error(Error::StsUnsupportedFormat, "Unsupported input type");
 
-    dst.create(xmap.size(), src.type());
+    _dst.create(xmap.size(), src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Scalar_<float> borderValueFloat;
     borderValueFloat = borderValue;

modules/gpuwarping/src/resize.cpp

@@ -44,7 +44,7 @@
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::resize(const GpuMat&, GpuMat&, Size, double, double, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::resize(InputArray, OutputArray, Size, double, double, int, Stream&) { throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
@@ -58,30 +58,34 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation, Stream& s)
+void cv::gpu::resize(InputArray _src, OutputArray _dst, Size dsize, double fx, double fy, int interpolation, Stream& _stream)
 {
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    GpuMat src = _src.getGpuMat();
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR
+
-            || interpolation == INTER_CUBIC || interpolation == INTER_AREA);
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
-    CV_Assert(!(dsize == Size()) || (fx > 0 && fy > 0));
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC || interpolation == INTER_AREA );
+    CV_Assert( !(dsize == Size()) || (fx > 0 && fy > 0) );
 
     if (dsize == Size())
+    {
         dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));
+    }
     else
     {
         fx = static_cast<double>(dsize.width) / src.cols;
         fy = static_cast<double>(dsize.height) / src.rows;
     }
-    if (dsize != dst.size())
+
-        dst.create(dsize, src.type());
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     if (dsize == src.size())
     {
-        src.copyTo(dst, s);
+        src.copyTo(dst, _stream);
         return;
     }
 
-    cudaStream_t stream = StreamAccessor::getStream(s);
+    cudaStream_t stream = StreamAccessor::getStream(_stream);
 
     Size wholeSize;
     Point ofs;

modules/gpuwarping/src/warp.cpp

@@ -47,17 +47,17 @@ using namespace cv::gpu;
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
+void cv::gpu::warpAffine(InputArray, OutputArray, InputArray, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
-void cv::gpu::buildWarpAffineMaps(const Mat&, bool, Size, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpAffineMaps(InputArray, bool, Size, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
+void cv::gpu::warpPerspective(InputArray, OutputArray, InputArray, Size, int, int, Scalar, Stream&) { throw_no_cuda(); }
-void cv::gpu::buildWarpPerspectiveMaps(const Mat&, bool, Size, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpPerspectiveMaps(InputArray, bool, Size, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::buildWarpPlaneMaps(Size, Rect, const Mat&, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpPlaneMaps(Size, Rect, InputArray, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
-void cv::gpu::buildWarpCylindricalMaps(Size, Rect, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpCylindricalMaps(Size, Rect, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
-void cv::gpu::buildWarpSphericalMaps(Size, Rect, const Mat&, const Mat&, float, GpuMat&, GpuMat&, Stream&) { throw_no_cuda(); }
+void cv::gpu::buildWarpSphericalMaps(Size, Rect, InputArray, InputArray, float, OutputArray, OutputArray, Stream&) { throw_no_cuda(); }
 
-void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int, Stream&) { throw_no_cuda(); }
+void cv::gpu::rotate(InputArray, OutputArray, Size, double, double, double, int, Stream&) { throw_no_cuda(); }
 
 #else // HAVE_CUDA
 
@@ -79,14 +79,19 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream)
+void cv::gpu::buildWarpAffineMaps(InputArray _M, bool inverse, Size dsize, OutputArray _xmap, OutputArray _ymap, Stream& stream)
 {
     using namespace cv::gpu::cudev::imgproc;
 
-    CV_Assert(M.rows == 2 && M.cols == 3);
+    Mat M = _M.getMat();
 
-    xmap.create(dsize, CV_32FC1);
+    CV_Assert( M.rows == 2 && M.cols == 3 );
-    ymap.create(dsize, CV_32FC1);
+
+    _xmap.create(dsize, CV_32FC1);
+    _ymap.create(dsize, CV_32FC1);
+
+    GpuMat xmap = _xmap.getGpuMat();
+    GpuMat ymap = _ymap.getGpuMat();
 
     float coeffs[2 * 3];
     Mat coeffsMat(2, 3, CV_32F, (void*)coeffs);
@@ -103,14 +108,19 @@ void cv::gpu::buildWarpAffineMaps(const Mat& M, bool inverse, Size dsize, GpuMat
     buildWarpAffineMaps_gpu(coeffs, xmap, ymap, StreamAccessor::getStream(stream));
 }
 
-void cv::gpu::buildWarpPerspectiveMaps(const Mat& M, bool inverse, Size dsize, GpuMat& xmap, GpuMat& ymap, Stream& stream)
+void cv::gpu::buildWarpPerspectiveMaps(InputArray _M, bool inverse, Size dsize, OutputArray _xmap, OutputArray _ymap, Stream& stream)
 {
     using namespace cv::gpu::cudev::imgproc;
 
-    CV_Assert(M.rows == 3 && M.cols == 3);
+    Mat M = _M.getMat();
 
-    xmap.create(dsize, CV_32FC1);
+    CV_Assert( M.rows == 3 && M.cols == 3 );
-    ymap.create(dsize, CV_32FC1);
+
+    _xmap.create(dsize, CV_32FC1);
+    _ymap.create(dsize, CV_32FC1);
+
+    GpuMat xmap = _xmap.getGpuMat();
+    GpuMat ymap = _ymap.getGpuMat();
 
     float coeffs[3 * 3];
     Mat coeffsMat(3, 3, CV_32F, (void*)coeffs);
@@ -174,17 +184,21 @@ namespace
     };
 }
 
-void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& s)
+void cv::gpu::warpAffine(InputArray _src, OutputArray _dst, InputArray _M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& stream)
 {
-    CV_Assert(M.rows == 2 && M.cols == 3);
+    GpuMat src = _src.getGpuMat();
+    Mat M = _M.getMat();
 
-    int interpolation = flags & INTER_MAX;
+    CV_Assert( M.rows == 2 && M.cols == 3 );
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    const int interpolation = flags & INTER_MAX;
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
 
-    dst.create(dsize, src.type());
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP );
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Size wholeSize;
     Point ofs;
@@ -258,7 +272,7 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
             }
         };
 
-        dst.setTo(borderValue);
+        dst.setTo(borderValue, stream);
 
         double coeffs[2][3];
         Mat coeffsMat(2, 3, CV_64F, (void*)coeffs);
@@ -267,7 +281,7 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
         const func_t func = funcs[(flags & WARP_INVERSE_MAP) != 0][src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(s));
+        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(stream));
     }
     else
     {
@@ -305,21 +319,25 @@ void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsiz
         borderValueFloat = borderValue;
 
         func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y, coeffs,
-            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(s), deviceSupports(FEATURE_SET_COMPUTE_20));
+            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
     }
 }
 
-void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& s)
+void cv::gpu::warpPerspective(InputArray _src, OutputArray _dst, InputArray _M, Size dsize, int flags, int borderMode, Scalar borderValue, Stream& stream)
 {
-    CV_Assert(M.rows == 3 && M.cols == 3);
+    GpuMat src = _src.getGpuMat();
+    Mat M = _M.getMat();
 
-    int interpolation = flags & INTER_MAX;
+    CV_Assert( M.rows == 3 && M.cols == 3 );
 
-    CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
+    const int interpolation = flags & INTER_MAX;
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
-    CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
 
-    dst.create(dsize, src.type());
+    CV_Assert( src.depth() <= CV_32F && src.channels() <= 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+    CV_Assert( borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP) ;
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
 
     Size wholeSize;
     Point ofs;
@@ -393,7 +411,7 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
             }
         };
 
-        dst.setTo(borderValue);
+        dst.setTo(borderValue, stream);
 
         double coeffs[3][3];
         Mat coeffsMat(3, 3, CV_64F, (void*)coeffs);
@@ -402,7 +420,7 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
         const func_t func = funcs[(flags & WARP_INVERSE_MAP) != 0][src.depth()][src.channels() - 1];
         CV_Assert(func != 0);
 
-        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(s));
+        func(src, dst, coeffs, interpolation, StreamAccessor::getStream(stream));
     }
     else
     {
@@ -440,7 +458,7 @@ void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size
         borderValueFloat = borderValue;
 
         func(src, PtrStepSzb(wholeSize.height, wholeSize.width, src.datastart, src.step), ofs.x, ofs.y, coeffs,
-            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(s), deviceSupports(FEATURE_SET_COMPUTE_20));
+            dst, interpolation, borderMode, borderValueFloat.val, StreamAccessor::getStream(stream), deviceSupports(FEATURE_SET_COMPUTE_20));
     }
 }
 
@@ -457,23 +475,30 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, const Mat &T,
+void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, InputArray _T,
-                                 float scale, GpuMat& map_x, GpuMat& map_y, Stream& stream)
+                                 float scale, OutputArray _map_x, OutputArray _map_y, Stream& stream)
 {
-    (void)src_size;
+    (void) src_size;
-    using namespace ::cv::gpu::cudev::imgproc;
 
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
+    Mat K = _K.getMat();
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
+    Mat R = _R.getMat();
-    CV_Assert((T.size() == Size(3,1) || T.size() == Size(1,3)) && T.type() == CV_32F && T.isContinuous());
+    Mat T = _T.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
+    CV_Assert( (T.size() == Size(3,1) || T.size() == Size(1,3)) && T.type() == CV_32FC1 && T.isContinuous() );
 
     Mat K_Rinv = K * R.t();
     Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
+    CV_Assert( K_Rinv.isContinuous() );
-    CV_Assert(R_Kinv.isContinuous());
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
 
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
     cudev::imgproc::buildWarpPlaneMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(),
                        T.ptr<float>(), scale, StreamAccessor::getStream(stream));
 }
@@ -491,22 +516,28 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
+void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, float scale,
-                                       GpuMat& map_x, GpuMat& map_y, Stream& stream)
+                                       OutputArray _map_x, OutputArray _map_y, Stream& stream)
 {
-    (void)src_size;
+    (void) src_size;
-    using namespace ::cv::gpu::cudev::imgproc;
 
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
+    Mat K = _K.getMat();
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
+    Mat R = _R.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
 
     Mat K_Rinv = K * R.t();
     Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
+    CV_Assert( K_Rinv.isContinuous() );
-    CV_Assert(R_Kinv.isContinuous());
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
 
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
     cudev::imgproc::buildWarpCylindricalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
 }
 
@@ -524,22 +555,28 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat &K, const Mat& R, float scale,
+void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, InputArray _K, InputArray _R, float scale,
-                                     GpuMat& map_x, GpuMat& map_y, Stream& stream)
+                                     OutputArray _map_x, OutputArray _map_y, Stream& stream)
 {
-    (void)src_size;
+    (void) src_size;
-    using namespace ::cv::gpu::cudev::imgproc;
 
-    CV_Assert(K.size() == Size(3,3) && K.type() == CV_32F);
+    Mat K = _K.getMat();
-    CV_Assert(R.size() == Size(3,3) && R.type() == CV_32F);
+    Mat R = _R.getMat();
+
+    CV_Assert( K.size() == Size(3,3) && K.type() == CV_32FC1 );
+    CV_Assert( R.size() == Size(3,3) && R.type() == CV_32FC1 );
 
     Mat K_Rinv = K * R.t();
     Mat R_Kinv = R * K.inv();
-    CV_Assert(K_Rinv.isContinuous());
+    CV_Assert( K_Rinv.isContinuous() );
-    CV_Assert(R_Kinv.isContinuous());
+    CV_Assert( R_Kinv.isContinuous() );
+
+    _map_x.create(dst_roi.size(), CV_32FC1);
+    _map_y.create(dst_roi.size(), CV_32FC1);
+
+    GpuMat map_x = _map_x.getGpuMat();
+    GpuMat map_y = _map_y.getGpuMat();
 
-    map_x.create(dst_roi.size(), CV_32F);
-    map_y.create(dst_roi.size(), CV_32F);
     cudev::imgproc::buildWarpSphericalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, K_Rinv.ptr<float>(), R_Kinv.ptr<float>(), scale, StreamAccessor::getStream(stream));
 }
 
@@ -589,10 +626,9 @@ namespace
     };
 }
 
-void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& stream)
+void cv::gpu::rotate(InputArray _src, OutputArray _dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& stream)
 {
     typedef void (*func_t)(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, cudaStream_t stream);
-
     static const func_t funcs[6][4] =
     {
         {NppRotate<CV_8U, nppiRotate_8u_C1R>::call, 0, NppRotate<CV_8U, nppiRotate_8u_C3R>::call, NppRotate<CV_8U, nppiRotate_8u_C4R>::call},
@@ -603,12 +639,16 @@ void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, d
         {NppRotate<CV_32F, nppiRotate_32f_C1R>::call, 0, NppRotate<CV_32F, nppiRotate_32f_C3R>::call, NppRotate<CV_32F, nppiRotate_32f_C4R>::call}
     };
 
-    CV_Assert(src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F);
+    GpuMat src = _src.getGpuMat();
-    CV_Assert(src.channels() == 1 || src.channels() == 3 || src.channels() == 4);
-    CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
 
-    dst.create(dsize, src.type());
+    CV_Assert( src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32F );
-    dst.setTo(Scalar::all(0));
+    CV_Assert( src.channels() == 1 || src.channels() == 3 || src.channels() == 4 );
+    CV_Assert( interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC );
+
+    _dst.create(dsize, src.type());
+    GpuMat dst = _dst.getGpuMat();
+
+    dst.setTo(Scalar::all(0), stream);
 
     funcs[src.depth()][src.channels() - 1](src, dst, dsize, angle, xShift, yShift, interpolation, StreamAccessor::getStream(stream));
 }

samples/gpu/cascadeclassifier.cpp

@@ -24,12 +24,11 @@ static void help()
 }
 
 
-template<class T>
+static void convertAndResize(const Mat& src, Mat& gray, Mat& resized, double scale)
-void convertAndResize(const T& src, T& gray, T& resized, double scale)
 {
     if (src.channels() == 3)
     {
-        cvtColor( src, gray, COLOR_BGR2GRAY );
+        cv::cvtColor( src, gray, COLOR_BGR2GRAY );
     }
     else
     {
@@ -40,7 +39,30 @@ void convertAndResize(const T& src, T& gray, T& resized, double scale)
 
     if (scale != 1)
     {
-        resize(gray, resized, sz);
+        cv::resize(gray, resized, sz);
+    }
+    else
+    {
+        resized = gray;
+    }
+}
+
+static void convertAndResize(const GpuMat& src, GpuMat& gray, GpuMat& resized, double scale)
+{
+    if (src.channels() == 3)
+    {
+        gpu::cvtColor( src, gray, COLOR_BGR2GRAY );
+    }
+    else
+    {
+        gray = src;
+    }
+
+    Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
+
+    if (scale != 1)
+    {
+        gpu::resize(gray, resized, sz);
     }
     else
     {