1 /*M///////////////////////////////////////////////////////////////////////////////////////
\r
3 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
\r
5 // By downloading, copying, installing or using the software you agree to this license.
\r
6 // If you do not agree to this license, do not download, install,
\r
7 // copy or use the software.
\r
10 // License Agreement
\r
11 // For Open Source Computer Vision Library
\r
13 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
\r
14 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
\r
15 // Third party copyrights are property of their respective owners.
\r
17 // Redistribution and use in source and binary forms, with or without modification,
\r
18 // are permitted provided that the following conditions are met:
\r
20 // * Redistribution's of source code must retain the above copyright notice,
\r
21 // this list of conditions and the following disclaimer.
\r
23 // * Redistribution's in binary form must reproduce the above copyright notice,
\r
24 // this list of conditions and the following disclaimer in the documentation
\r
25 // and/or other materials provided with the distribution.
\r
27 // * The name of the copyright holders may not be used to endorse or promote products
\r
28 // derived from this software without specific prior written permission.
\r
30 // This software is provided by the copyright holders and contributors "as is" and
\r
31 // any express or implied warranties, including, but not limited to, the implied
\r
32 // warranties of merchantability and fitness for a particular purpose are disclaimed.
\r
33 // In no event shall the Intel Corporation or contributors be liable for any direct,
\r
34 // indirect, incidental, special, exemplary, or consequential damages
\r
35 // (including, but not limited to, procurement of substitute goods or services;
\r
36 // loss of use, data, or profits; or business interruption) however caused
\r
37 // and on any theory of liability, whether in contract, strict liability,
\r
38 // or tort (including negligence or otherwise) arising in any way out of
\r
39 // the use of this software, even if advised of the possibility of such damage.
\r
43 #include "precomp.hpp"
\r
46 using namespace cv::gpu;
\r
48 #if !defined (HAVE_CUDA)
\r
50 void cv::gpu::remap(const GpuMat&, GpuMat&, const GpuMat&, const GpuMat&){ throw_nogpu(); }
\r
51 void cv::gpu::meanShiftFiltering(const GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
\r
52 void cv::gpu::meanShiftProc(const GpuMat&, GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
\r
53 void cv::gpu::drawColorDisp(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
\r
54 void cv::gpu::reprojectImageTo3D(const GpuMat&, GpuMat&, const Mat&, Stream&) { throw_nogpu(); }
\r
55 void cv::gpu::resize(const GpuMat&, GpuMat&, Size, double, double, int, Stream&) { throw_nogpu(); }
\r
56 void cv::gpu::copyMakeBorder(const GpuMat&, GpuMat&, int, int, int, int, const Scalar&, Stream&) { throw_nogpu(); }
\r
57 void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int, Stream&) { throw_nogpu(); }
\r
58 void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int, Stream&) { throw_nogpu(); }
\r
59 void cv::gpu::buildWarpPlaneMaps(Size, Rect, const Mat&, double, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
60 void cv::gpu::buildWarpCylindricalMaps(Size, Rect, const Mat&, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
61 void cv::gpu::buildWarpSphericalMaps(Size, Rect, const Mat&, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
62 void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int, Stream&) { throw_nogpu(); }
\r
63 void cv::gpu::integral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
64 void cv::gpu::integralBuffered(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
65 void cv::gpu::integral(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
66 void cv::gpu::sqrIntegral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
67 void cv::gpu::columnSum(const GpuMat&, GpuMat&) { throw_nogpu(); }
\r
68 void cv::gpu::rectStdDev(const GpuMat&, const GpuMat&, GpuMat&, const Rect&, Stream&) { throw_nogpu(); }
\r
69 void cv::gpu::evenLevels(GpuMat&, int, int, int) { throw_nogpu(); }
\r
70 void cv::gpu::histEven(const GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
\r
71 void cv::gpu::histEven(const GpuMat&, GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
\r
72 void cv::gpu::histEven(const GpuMat&, GpuMat*, int*, int*, int*, Stream&) { throw_nogpu(); }
\r
73 void cv::gpu::histEven(const GpuMat&, GpuMat*, GpuMat&, int*, int*, int*, Stream&) { throw_nogpu(); }
\r
74 void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
\r
75 void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
76 void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, Stream&) { throw_nogpu(); }
\r
77 void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, GpuMat&, Stream&) { throw_nogpu(); }
\r
78 void cv::gpu::calcHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
79 void cv::gpu::calcHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
80 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
81 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
82 void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
83 void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
\r
84 void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
\r
85 void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
\r
86 void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
\r
87 void cv::gpu::mulSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, bool) { throw_nogpu(); }
\r
88 void cv::gpu::mulAndScaleSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, float, bool) { throw_nogpu(); }
\r
89 void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int) { throw_nogpu(); }
\r
90 void cv::gpu::ConvolveBuf::create(Size, Size) { throw_nogpu(); }
\r
91 void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool) { throw_nogpu(); }
\r
92 void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&) { throw_nogpu(); }
\r
93 void cv::gpu::downsample(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
94 void cv::gpu::upsample(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
95 void cv::gpu::pyrDown(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
96 void cv::gpu::PyrDownBuf::create(Size, int) { throw_nogpu(); }
\r
97 void cv::gpu::pyrDown(const GpuMat&, GpuMat&, PyrDownBuf&, Stream&) { throw_nogpu(); }
\r
98 void cv::gpu::pyrUp(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
\r
99 void cv::gpu::PyrUpBuf::create(Size, int) { throw_nogpu(); }
\r
100 void cv::gpu::pyrUp(const GpuMat&, GpuMat&, PyrUpBuf&, Stream&) { throw_nogpu(); }
\r
101 void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
\r
102 void cv::gpu::Canny(const GpuMat&, CannyBuf&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
\r
103 void cv::gpu::Canny(const GpuMat&, const GpuMat&, GpuMat&, double, double, bool) { throw_nogpu(); }
\r
104 void cv::gpu::Canny(const GpuMat&, const GpuMat&, CannyBuf&, GpuMat&, double, double, bool) { throw_nogpu(); }
\r
105 cv::gpu::CannyBuf::CannyBuf(const GpuMat&, const GpuMat&) { throw_nogpu(); }
\r
106 void cv::gpu::CannyBuf::create(const Size&, int) { throw_nogpu(); }
\r
107 void cv::gpu::CannyBuf::release() { throw_nogpu(); }
\r
109 #else /* !defined (HAVE_CUDA) */
\r
111 namespace cv { namespace gpu { namespace imgproc
\r
113 void remap_gpu_1c(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);
\r
114 void remap_gpu_3c(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);
\r
116 extern "C" void meanShiftFiltering_gpu(const DevMem2D& src, DevMem2D dst, int sp, int sr, int maxIter, float eps);
\r
117 extern "C" void meanShiftProc_gpu(const DevMem2D& src, DevMem2D dstr, DevMem2D dstsp, int sp, int sr, int maxIter, float eps);
\r
119 void drawColorDisp_gpu(const DevMem2D& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);
\r
120 void drawColorDisp_gpu(const DevMem2D_<short>& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);
\r
122 void reprojectImageTo3D_gpu(const DevMem2D& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
\r
123 void reprojectImageTo3D_gpu(const DevMem2D_<short>& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
\r
126 ////////////////////////////////////////////////////////////////////////
\r
129 void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap)
\r
131 typedef void (*remap_gpu_t)(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, DevMem2D dst);
\r
132 static const remap_gpu_t callers[] = {imgproc::remap_gpu_1c, 0, imgproc::remap_gpu_3c};
\r
134 CV_Assert((src.type() == CV_8U || src.type() == CV_8UC3) && xmap.type() == CV_32F && ymap.type() == CV_32F);
\r
136 dst.create(xmap.size(), src.type());
\r
138 callers[src.channels() - 1](src, xmap, ymap, dst);
\r
141 ////////////////////////////////////////////////////////////////////////
\r
142 // meanShiftFiltering_GPU
\r
144 void cv::gpu::meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria)
\r
147 CV_Error( CV_StsBadArg, "The input image is empty" );
\r
149 if( src.depth() != CV_8U || src.channels() != 4 )
\r
150 CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
\r
152 dst.create( src.size(), CV_8UC4 );
\r
154 if( !(criteria.type & TermCriteria::MAX_ITER) )
\r
155 criteria.maxCount = 5;
\r
157 int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
\r
160 if( !(criteria.type & TermCriteria::EPS) )
\r
162 eps = (float)std::max(criteria.epsilon, 0.0);
\r
164 imgproc::meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps);
\r
167 ////////////////////////////////////////////////////////////////////////
\r
168 // meanShiftProc_GPU
\r
170 void cv::gpu::meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria)
\r
173 CV_Error( CV_StsBadArg, "The input image is empty" );
\r
175 if( src.depth() != CV_8U || src.channels() != 4 )
\r
176 CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
\r
178 dstr.create( src.size(), CV_8UC4 );
\r
179 dstsp.create( src.size(), CV_16SC2 );
\r
181 if( !(criteria.type & TermCriteria::MAX_ITER) )
\r
182 criteria.maxCount = 5;
\r
184 int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
\r
187 if( !(criteria.type & TermCriteria::EPS) )
\r
189 eps = (float)std::max(criteria.epsilon, 0.0);
\r
191 imgproc::meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps);
\r
194 ////////////////////////////////////////////////////////////////////////
\r
199 template <typename T>
\r
200 void drawColorDisp_caller(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream)
\r
202 dst.create(src.size(), CV_8UC4);
\r
204 imgproc::drawColorDisp_gpu((DevMem2D_<T>)src, dst, ndisp, stream);
\r
207 typedef void (*drawColorDisp_caller_t)(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream);
\r
209 const drawColorDisp_caller_t drawColorDisp_callers[] = {drawColorDisp_caller<unsigned char>, 0, 0, drawColorDisp_caller<short>, 0, 0, 0, 0};
\r
212 void cv::gpu::drawColorDisp(const GpuMat& src, GpuMat& dst, int ndisp, Stream& stream)
\r
214 CV_Assert(src.type() == CV_8U || src.type() == CV_16S);
\r
216 drawColorDisp_callers[src.type()](src, dst, ndisp, StreamAccessor::getStream(stream));
\r
219 ////////////////////////////////////////////////////////////////////////
\r
220 // reprojectImageTo3D
\r
224 template <typename T>
\r
225 void reprojectImageTo3D_caller(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream)
\r
227 xyzw.create(disp.rows, disp.cols, CV_32FC4);
\r
228 imgproc::reprojectImageTo3D_gpu((DevMem2D_<T>)disp, xyzw, Q.ptr<float>(), stream);
\r
231 typedef void (*reprojectImageTo3D_caller_t)(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream);
\r
233 const reprojectImageTo3D_caller_t reprojectImageTo3D_callers[] = {reprojectImageTo3D_caller<unsigned char>, 0, 0, reprojectImageTo3D_caller<short>, 0, 0, 0, 0};
\r
236 void cv::gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, Stream& stream)
\r
238 CV_Assert((disp.type() == CV_8U || disp.type() == CV_16S) && Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4);
\r
240 reprojectImageTo3D_callers[disp.type()](disp, xyzw, Q, StreamAccessor::getStream(stream));
\r
243 ////////////////////////////////////////////////////////////////////////
\r
246 void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation, Stream& s)
\r
248 static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR/*, NPPI_INTER_CUBIC, 0, NPPI_INTER_LANCZOS*/};
\r
250 CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
\r
251 CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR/* || interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4*/);
\r
253 CV_Assert( src.size().area() > 0 );
\r
254 CV_Assert( !(dsize == Size()) || (fx > 0 && fy > 0) );
\r
256 if( dsize == Size() )
\r
258 dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));
\r
262 fx = (double)dsize.width / src.cols;
\r
263 fy = (double)dsize.height / src.rows;
\r
266 dst.create(dsize, src.type());
\r
269 srcsz.width = src.cols;
\r
270 srcsz.height = src.rows;
\r
272 srcrect.x = srcrect.y = 0;
\r
273 srcrect.width = src.cols;
\r
274 srcrect.height = src.rows;
\r
276 dstsz.width = dst.cols;
\r
277 dstsz.height = dst.rows;
\r
279 cudaStream_t stream = StreamAccessor::getStream(s);
\r
281 NppStreamHandler h(stream);
\r
283 if (src.type() == CV_8UC1)
\r
285 nppSafeCall( nppiResize_8u_C1R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcrect,
\r
286 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, fx, fy, npp_inter[interpolation]) );
\r
290 nppSafeCall( nppiResize_8u_C4R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcrect,
\r
291 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, fx, fy, npp_inter[interpolation]) );
\r
295 cudaSafeCall( cudaDeviceSynchronize() );
\r
298 ////////////////////////////////////////////////////////////////////////
\r
301 void cv::gpu::copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, const Scalar& value, Stream& s)
\r
303 CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4 || src.type() == CV_32SC1 || src.type() == CV_32FC1);
\r
305 dst.create(src.rows + top + bottom, src.cols + left + right, src.type());
\r
308 srcsz.width = src.cols;
\r
309 srcsz.height = src.rows;
\r
311 dstsz.width = dst.cols;
\r
312 dstsz.height = dst.rows;
\r
314 cudaStream_t stream = StreamAccessor::getStream(s);
\r
316 NppStreamHandler h(stream);
\r
318 switch (src.type())
\r
322 Npp8u nVal = static_cast<Npp8u>(value[0]);
\r
323 nppSafeCall( nppiCopyConstBorder_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
\r
324 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
\r
329 Npp8u nVal[] = {static_cast<Npp8u>(value[0]), static_cast<Npp8u>(value[1]), static_cast<Npp8u>(value[2]), static_cast<Npp8u>(value[3])};
\r
330 nppSafeCall( nppiCopyConstBorder_8u_C4R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
\r
331 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
\r
336 Npp32s nVal = static_cast<Npp32s>(value[0]);
\r
337 nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
\r
338 dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
\r
343 Npp32f val = static_cast<Npp32f>(value[0]);
\r
344 Npp32s nVal = *(reinterpret_cast<Npp32s*>(&val));
\r
345 nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
\r
346 dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
\r
350 CV_Assert(!"Unsupported source type");
\r
354 cudaSafeCall( cudaDeviceSynchronize() );
\r
357 ////////////////////////////////////////////////////////////////////////
\r
362 typedef NppStatus (*npp_warp_8u_t)(const Npp8u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp8u* pDst,
\r
363 int dstStep, NppiRect dstRoi, const double coeffs[][3],
\r
364 int interpolation);
\r
365 typedef NppStatus (*npp_warp_16u_t)(const Npp16u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp16u* pDst,
\r
366 int dstStep, NppiRect dstRoi, const double coeffs[][3],
\r
367 int interpolation);
\r
368 typedef NppStatus (*npp_warp_32s_t)(const Npp32s* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32s* pDst,
\r
369 int dstStep, NppiRect dstRoi, const double coeffs[][3],
\r
370 int interpolation);
\r
371 typedef NppStatus (*npp_warp_32f_t)(const Npp32f* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32f* pDst,
\r
372 int dstStep, NppiRect dstRoi, const double coeffs[][3],
\r
373 int interpolation);
\r
375 void nppWarpCaller(const GpuMat& src, GpuMat& dst, double coeffs[][3], const Size& dsize, int flags,
\r
376 npp_warp_8u_t npp_warp_8u[][2], npp_warp_16u_t npp_warp_16u[][2],
\r
377 npp_warp_32s_t npp_warp_32s[][2], npp_warp_32f_t npp_warp_32f[][2], cudaStream_t stream)
\r
379 static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
\r
381 int interpolation = flags & INTER_MAX;
\r
383 CV_Assert((src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S || src.depth() == CV_32F) && src.channels() != 2);
\r
384 CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
\r
386 dst.create(dsize, src.type());
\r
389 srcsz.height = src.rows;
\r
390 srcsz.width = src.cols;
\r
392 srcroi.x = srcroi.y = 0;
\r
393 srcroi.height = src.rows;
\r
394 srcroi.width = src.cols;
\r
396 dstroi.x = dstroi.y = 0;
\r
397 dstroi.height = dst.rows;
\r
398 dstroi.width = dst.cols;
\r
400 int warpInd = (flags & WARP_INVERSE_MAP) >> 4;
\r
402 NppStreamHandler h(stream);
\r
404 switch (src.depth())
\r
407 nppSafeCall( npp_warp_8u[src.channels()][warpInd](src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
\r
408 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
\r
411 nppSafeCall( npp_warp_16u[src.channels()][warpInd](src.ptr<Npp16u>(), srcsz, static_cast<int>(src.step), srcroi,
\r
412 dst.ptr<Npp16u>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
\r
415 nppSafeCall( npp_warp_32s[src.channels()][warpInd](src.ptr<Npp32s>(), srcsz, static_cast<int>(src.step), srcroi,
\r
416 dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
\r
419 nppSafeCall( npp_warp_32f[src.channels()][warpInd](src.ptr<Npp32f>(), srcsz, static_cast<int>(src.step), srcroi,
\r
420 dst.ptr<Npp32f>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
\r
423 CV_Assert(!"Unsupported source type");
\r
427 cudaSafeCall( cudaDeviceSynchronize() );
\r
431 void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, Stream& s)
\r
433 static npp_warp_8u_t npp_warpAffine_8u[][2] =
\r
436 {nppiWarpAffine_8u_C1R, nppiWarpAffineBack_8u_C1R},
\r
438 {nppiWarpAffine_8u_C3R, nppiWarpAffineBack_8u_C3R},
\r
439 {nppiWarpAffine_8u_C4R, nppiWarpAffineBack_8u_C4R}
\r
441 static npp_warp_16u_t npp_warpAffine_16u[][2] =
\r
444 {nppiWarpAffine_16u_C1R, nppiWarpAffineBack_16u_C1R},
\r
446 {nppiWarpAffine_16u_C3R, nppiWarpAffineBack_16u_C3R},
\r
447 {nppiWarpAffine_16u_C4R, nppiWarpAffineBack_16u_C4R}
\r
449 static npp_warp_32s_t npp_warpAffine_32s[][2] =
\r
452 {nppiWarpAffine_32s_C1R, nppiWarpAffineBack_32s_C1R},
\r
454 {nppiWarpAffine_32s_C3R, nppiWarpAffineBack_32s_C3R},
\r
455 {nppiWarpAffine_32s_C4R, nppiWarpAffineBack_32s_C4R}
\r
457 static npp_warp_32f_t npp_warpAffine_32f[][2] =
\r
460 {nppiWarpAffine_32f_C1R, nppiWarpAffineBack_32f_C1R},
\r
462 {nppiWarpAffine_32f_C3R, nppiWarpAffineBack_32f_C3R},
\r
463 {nppiWarpAffine_32f_C4R, nppiWarpAffineBack_32f_C4R}
\r
466 CV_Assert(M.rows == 2 && M.cols == 3);
\r
468 double coeffs[2][3];
\r
469 Mat coeffsMat(2, 3, CV_64F, (void*)coeffs);
\r
470 M.convertTo(coeffsMat, coeffsMat.type());
\r
472 nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpAffine_8u, npp_warpAffine_16u, npp_warpAffine_32s, npp_warpAffine_32f, StreamAccessor::getStream(s));
\r
475 void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, Stream& s)
\r
477 static npp_warp_8u_t npp_warpPerspective_8u[][2] =
\r
480 {nppiWarpPerspective_8u_C1R, nppiWarpPerspectiveBack_8u_C1R},
\r
482 {nppiWarpPerspective_8u_C3R, nppiWarpPerspectiveBack_8u_C3R},
\r
483 {nppiWarpPerspective_8u_C4R, nppiWarpPerspectiveBack_8u_C4R}
\r
485 static npp_warp_16u_t npp_warpPerspective_16u[][2] =
\r
488 {nppiWarpPerspective_16u_C1R, nppiWarpPerspectiveBack_16u_C1R},
\r
490 {nppiWarpPerspective_16u_C3R, nppiWarpPerspectiveBack_16u_C3R},
\r
491 {nppiWarpPerspective_16u_C4R, nppiWarpPerspectiveBack_16u_C4R}
\r
493 static npp_warp_32s_t npp_warpPerspective_32s[][2] =
\r
496 {nppiWarpPerspective_32s_C1R, nppiWarpPerspectiveBack_32s_C1R},
\r
498 {nppiWarpPerspective_32s_C3R, nppiWarpPerspectiveBack_32s_C3R},
\r
499 {nppiWarpPerspective_32s_C4R, nppiWarpPerspectiveBack_32s_C4R}
\r
501 static npp_warp_32f_t npp_warpPerspective_32f[][2] =
\r
504 {nppiWarpPerspective_32f_C1R, nppiWarpPerspectiveBack_32f_C1R},
\r
506 {nppiWarpPerspective_32f_C3R, nppiWarpPerspectiveBack_32f_C3R},
\r
507 {nppiWarpPerspective_32f_C4R, nppiWarpPerspectiveBack_32f_C4R}
\r
510 CV_Assert(M.rows == 3 && M.cols == 3);
\r
512 double coeffs[3][3];
\r
513 Mat coeffsMat(3, 3, CV_64F, (void*)coeffs);
\r
514 M.convertTo(coeffsMat, coeffsMat.type());
\r
516 nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpPerspective_8u, npp_warpPerspective_16u, npp_warpPerspective_32s, npp_warpPerspective_32f, StreamAccessor::getStream(s));
\r
519 //////////////////////////////////////////////////////////////////////////////
\r
520 // buildWarpPlaneMaps
\r
522 namespace cv { namespace gpu { namespace imgproc
\r
524 void buildWarpPlaneMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
\r
525 const float r[9], const float rinv[9], float f, float s, float dist,
\r
526 float half_w, float half_h, cudaStream_t stream);
\r
529 void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
\r
530 double dist, GpuMat& map_x, GpuMat& map_y, Stream& stream)
\r
532 CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
\r
533 Mat Rinv = R.inv();
\r
534 CV_Assert(Rinv.isContinuous());
\r
536 map_x.create(dst_roi.size(), CV_32F);
\r
537 map_y.create(dst_roi.size(), CV_32F);
\r
538 imgproc::buildWarpPlaneMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
\r
539 static_cast<float>(f), static_cast<float>(s), static_cast<float>(dist),
\r
540 0.5f*src_size.width, 0.5f*src_size.height, StreamAccessor::getStream(stream));
\r
543 //////////////////////////////////////////////////////////////////////////////
\r
544 // buildWarpCylyndricalMaps
\r
546 namespace cv { namespace gpu { namespace imgproc
\r
548 void buildWarpCylindricalMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
\r
549 const float r[9], const float rinv[9], float f, float s,
\r
550 float half_w, float half_h, cudaStream_t stream);
\r
553 void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
\r
554 GpuMat& map_x, GpuMat& map_y, Stream& stream)
\r
556 CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
\r
557 Mat Rinv = R.inv();
\r
558 CV_Assert(Rinv.isContinuous());
\r
560 map_x.create(dst_roi.size(), CV_32F);
\r
561 map_y.create(dst_roi.size(), CV_32F);
\r
562 imgproc::buildWarpCylindricalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
\r
563 static_cast<float>(f), static_cast<float>(s), 0.5f*src_size.width, 0.5f*src_size.height,
\r
564 StreamAccessor::getStream(stream));
\r
568 //////////////////////////////////////////////////////////////////////////////
\r
569 // buildWarpSphericalMaps
\r
571 namespace cv { namespace gpu { namespace imgproc
\r
573 void buildWarpSphericalMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
\r
574 const float r[9], const float rinv[9], float f, float s,
\r
575 float half_w, float half_h, cudaStream_t stream);
\r
578 void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
\r
579 GpuMat& map_x, GpuMat& map_y, Stream& stream)
\r
581 CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
\r
582 Mat Rinv = R.inv();
\r
583 CV_Assert(Rinv.isContinuous());
\r
585 map_x.create(dst_roi.size(), CV_32F);
\r
586 map_y.create(dst_roi.size(), CV_32F);
\r
587 imgproc::buildWarpSphericalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
\r
588 static_cast<float>(f), static_cast<float>(s), 0.5f*src_size.width, 0.5f*src_size.height,
\r
589 StreamAccessor::getStream(stream));
\r
592 ////////////////////////////////////////////////////////////////////////
\r
595 void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& s)
\r
597 static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
\r
599 CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
\r
600 CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
\r
602 dst.create(dsize, src.type());
\r
605 srcsz.height = src.rows;
\r
606 srcsz.width = src.cols;
\r
608 srcroi.x = srcroi.y = 0;
\r
609 srcroi.height = src.rows;
\r
610 srcroi.width = src.cols;
\r
612 dstroi.x = dstroi.y = 0;
\r
613 dstroi.height = dst.rows;
\r
614 dstroi.width = dst.cols;
\r
616 cudaStream_t stream = StreamAccessor::getStream(s);
\r
618 NppStreamHandler h(stream);
\r
620 if (src.type() == CV_8UC1)
\r
622 nppSafeCall( nppiRotate_8u_C1R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
\r
623 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
\r
627 nppSafeCall( nppiRotate_8u_C4R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
\r
628 dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
\r
632 cudaSafeCall( cudaDeviceSynchronize() );
\r
635 ////////////////////////////////////////////////////////////////////////
\r
638 void cv::gpu::integral(const GpuMat& src, GpuMat& sum, Stream& s)
\r
641 integralBuffered(src, sum, buffer, s);
\r
644 void cv::gpu::integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer, Stream& s)
\r
646 CV_Assert(src.type() == CV_8UC1);
\r
648 sum.create(src.rows + 1, src.cols + 1, CV_32S);
\r
650 NcvSize32u roiSize;
\r
651 roiSize.width = src.cols;
\r
652 roiSize.height = src.rows;
\r
654 cudaDeviceProp prop;
\r
655 cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
\r
658 nppSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize, prop) );
\r
659 ensureSizeIsEnough(1, bufSize, CV_8UC1, buffer);
\r
661 cudaStream_t stream = StreamAccessor::getStream(s);
\r
663 NppStStreamHandler h(stream);
\r
665 nppSafeCall( nppiStIntegral_8u32u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>()), static_cast<int>(src.step),
\r
666 sum.ptr<Ncv32u>(), static_cast<int>(sum.step), roiSize, buffer.ptr<Ncv8u>(), bufSize, prop) );
\r
669 cudaSafeCall( cudaDeviceSynchronize() );
\r
672 void cv::gpu::integral(const GpuMat& src, GpuMat& sum, GpuMat& sqsum, Stream& s)
\r
674 CV_Assert(src.type() == CV_8UC1);
\r
676 int width = src.cols + 1, height = src.rows + 1;
\r
678 sum.create(height, width, CV_32S);
\r
679 sqsum.create(height, width, CV_32F);
\r
682 sz.width = src.cols;
\r
683 sz.height = src.rows;
\r
685 cudaStream_t stream = StreamAccessor::getStream(s);
\r
687 NppStreamHandler h(stream);
\r
689 nppSafeCall( nppiSqrIntegral_8u32s32f_C1R(const_cast<Npp8u*>(src.ptr<Npp8u>()), static_cast<int>(src.step),
\r
690 sum.ptr<Npp32s>(), static_cast<int>(sum.step), sqsum.ptr<Npp32f>(), static_cast<int>(sqsum.step), sz, 0, 0.0f, height) );
\r
693 cudaSafeCall( cudaDeviceSynchronize() );
\r
696 //////////////////////////////////////////////////////////////////////////////
\r
699 void cv::gpu::sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& s)
\r
701 CV_Assert(src.type() == CV_8U);
\r
703 NcvSize32u roiSize;
\r
704 roiSize.width = src.cols;
\r
705 roiSize.height = src.rows;
\r
707 cudaDeviceProp prop;
\r
708 cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
\r
711 nppSafeCall(nppiStSqrIntegralGetSize_8u64u(roiSize, &bufSize, prop));
\r
712 GpuMat buf(1, bufSize, CV_8U);
\r
714 cudaStream_t stream = StreamAccessor::getStream(s);
\r
716 NppStStreamHandler h(stream);
\r
718 sqsum.create(src.rows + 1, src.cols + 1, CV_64F);
\r
719 nppSafeCall(nppiStSqrIntegral_8u64u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>(0)), static_cast<int>(src.step),
\r
720 sqsum.ptr<Ncv64u>(0), static_cast<int>(sqsum.step), roiSize, buf.ptr<Ncv8u>(0), bufSize, prop));
\r
723 cudaSafeCall( cudaDeviceSynchronize() );
\r
726 //////////////////////////////////////////////////////////////////////////////
\r
729 namespace cv { namespace gpu { namespace imgproc
\r
731 void columnSum_32F(const DevMem2D src, const DevMem2D dst);
\r
734 void cv::gpu::columnSum(const GpuMat& src, GpuMat& dst)
\r
736 CV_Assert(src.type() == CV_32F);
\r
738 dst.create(src.size(), CV_32F);
\r
739 imgproc::columnSum_32F(src, dst);
\r
742 void cv::gpu::rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& s)
\r
744 CV_Assert(src.type() == CV_32SC1 && sqr.type() == CV_32FC1);
\r
746 dst.create(src.size(), CV_32FC1);
\r
749 sz.width = src.cols;
\r
750 sz.height = src.rows;
\r
753 nppRect.height = rect.height;
\r
754 nppRect.width = rect.width;
\r
755 nppRect.x = rect.x;
\r
756 nppRect.y = rect.y;
\r
758 cudaStream_t stream = StreamAccessor::getStream(s);
\r
760 NppStreamHandler h(stream);
\r
762 nppSafeCall( nppiRectStdDev_32s32f_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), sqr.ptr<Npp32f>(), static_cast<int>(sqr.step),
\r
763 dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, nppRect) );
\r
766 cudaSafeCall( cudaDeviceSynchronize() );
\r
770 ////////////////////////////////////////////////////////////////////////
\r
775 template<int n> struct NPPTypeTraits;
\r
776 template<> struct NPPTypeTraits<CV_8U> { typedef Npp8u npp_type; };
\r
777 template<> struct NPPTypeTraits<CV_16U> { typedef Npp16u npp_type; };
\r
778 template<> struct NPPTypeTraits<CV_16S> { typedef Npp16s npp_type; };
\r
779 template<> struct NPPTypeTraits<CV_32F> { typedef Npp32f npp_type; };
\r
781 typedef NppStatus (*get_buf_size_c1_t)(NppiSize oSizeROI, int nLevels, int* hpBufferSize);
\r
782 typedef NppStatus (*get_buf_size_c4_t)(NppiSize oSizeROI, int nLevels[], int* hpBufferSize);
\r
784 template<int SDEPTH> struct NppHistogramEvenFuncC1
\r
786 typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
\r
788 typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s * pHist,
\r
789 int nLevels, Npp32s nLowerLevel, Npp32s nUpperLevel, Npp8u * pBuffer);
\r
791 template<int SDEPTH> struct NppHistogramEvenFuncC4
\r
793 typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
\r
795 typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI,
\r
796 Npp32s * pHist[4], int nLevels[4], Npp32s nLowerLevel[4], Npp32s nUpperLevel[4], Npp8u * pBuffer);
\r
799 template<int SDEPTH, typename NppHistogramEvenFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
\r
800 struct NppHistogramEvenC1
\r
802 typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;
\r
804 static void hist(const GpuMat& src, GpuMat& hist, GpuMat& buffer, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)
\r
806 int levels = histSize + 1;
\r
807 hist.create(1, histSize, CV_32S);
\r
810 sz.width = src.cols;
\r
811 sz.height = src.rows;
\r
814 get_buf_size(sz, levels, &buf_size);
\r
816 ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
\r
818 NppStreamHandler h(stream);
\r
820 nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels,
\r
821 lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
\r
824 cudaSafeCall( cudaDeviceSynchronize() );
\r
827 template<int SDEPTH, typename NppHistogramEvenFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
\r
828 struct NppHistogramEvenC4
\r
830 typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;
\r
832 static void hist(const GpuMat& src, GpuMat hist[4], GpuMat& buffer, int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)
\r
834 int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};
\r
835 hist[0].create(1, histSize[0], CV_32S);
\r
836 hist[1].create(1, histSize[1], CV_32S);
\r
837 hist[2].create(1, histSize[2], CV_32S);
\r
838 hist[3].create(1, histSize[3], CV_32S);
\r
841 sz.width = src.cols;
\r
842 sz.height = src.rows;
\r
844 Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
\r
847 get_buf_size(sz, levels, &buf_size);
\r
849 ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
\r
851 NppStreamHandler h(stream);
\r
853 nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, levels, lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
\r
856 cudaSafeCall( cudaDeviceSynchronize() );
\r
860 template<int SDEPTH> struct NppHistogramRangeFuncC1
\r
862 typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
\r
863 typedef Npp32s level_t;
\r
864 enum {LEVEL_TYPE_CODE=CV_32SC1};
\r
866 typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
\r
867 const Npp32s* pLevels, int nLevels, Npp8u* pBuffer);
\r
869 template<> struct NppHistogramRangeFuncC1<CV_32F>
\r
871 typedef Npp32f src_t;
\r
872 typedef Npp32f level_t;
\r
873 enum {LEVEL_TYPE_CODE=CV_32FC1};
\r
875 typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
\r
876 const Npp32f* pLevels, int nLevels, Npp8u* pBuffer);
\r
878 template<int SDEPTH> struct NppHistogramRangeFuncC4
\r
880 typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
\r
881 typedef Npp32s level_t;
\r
882 enum {LEVEL_TYPE_CODE=CV_32SC1};
\r
884 typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
\r
885 const Npp32s* pLevels[4], int nLevels[4], Npp8u* pBuffer);
\r
887 template<> struct NppHistogramRangeFuncC4<CV_32F>
\r
889 typedef Npp32f src_t;
\r
890 typedef Npp32f level_t;
\r
891 enum {LEVEL_TYPE_CODE=CV_32FC1};
\r
893 typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
\r
894 const Npp32f* pLevels[4], int nLevels[4], Npp8u* pBuffer);
\r
897 template<int SDEPTH, typename NppHistogramRangeFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
\r
898 struct NppHistogramRangeC1
\r
900 typedef typename NppHistogramRangeFuncC1<SDEPTH>::src_t src_t;
\r
901 typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
\r
902 enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
\r
904 static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buffer, cudaStream_t stream)
\r
906 CV_Assert(levels.type() == LEVEL_TYPE_CODE && levels.rows == 1);
\r
908 hist.create(1, levels.cols - 1, CV_32S);
\r
911 sz.width = src.cols;
\r
912 sz.height = src.rows;
\r
915 get_buf_size(sz, levels.cols, &buf_size);
\r
917 ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
\r
919 NppStreamHandler h(stream);
\r
921 nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels.ptr<level_t>(), levels.cols, buffer.ptr<Npp8u>()) );
\r
924 cudaSafeCall( cudaDeviceSynchronize() );
\r
927 template<int SDEPTH, typename NppHistogramRangeFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
\r
928 struct NppHistogramRangeC4
\r
930 typedef typename NppHistogramRangeFuncC4<SDEPTH>::src_t src_t;
\r
931 typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
\r
932 enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
\r
934 static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buffer, cudaStream_t stream)
\r
936 CV_Assert(levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1);
\r
937 CV_Assert(levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1);
\r
938 CV_Assert(levels[2].type() == LEVEL_TYPE_CODE && levels[2].rows == 1);
\r
939 CV_Assert(levels[3].type() == LEVEL_TYPE_CODE && levels[3].rows == 1);
\r
941 hist[0].create(1, levels[0].cols - 1, CV_32S);
\r
942 hist[1].create(1, levels[1].cols - 1, CV_32S);
\r
943 hist[2].create(1, levels[2].cols - 1, CV_32S);
\r
944 hist[3].create(1, levels[3].cols - 1, CV_32S);
\r
946 Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
\r
947 int nLevels[] = {levels[0].cols, levels[1].cols, levels[2].cols, levels[3].cols};
\r
948 const level_t* pLevels[] = {levels[0].ptr<level_t>(), levels[1].ptr<level_t>(), levels[2].ptr<level_t>(), levels[3].ptr<level_t>()};
\r
951 sz.width = src.cols;
\r
952 sz.height = src.rows;
\r
955 get_buf_size(sz, nLevels, &buf_size);
\r
957 ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
\r
959 NppStreamHandler h(stream);
\r
961 nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, pLevels, nLevels, buffer.ptr<Npp8u>()) );
\r
964 cudaSafeCall( cudaDeviceSynchronize() );
\r
969 void cv::gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel)
\r
971 Mat host_levels(1, nLevels, CV_32SC1);
\r
972 nppSafeCall( nppiEvenLevelsHost_32s(host_levels.ptr<Npp32s>(), nLevels, lowerLevel, upperLevel) );
\r
973 levels.upload(host_levels);
\r
976 void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream)
\r
979 histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
\r
982 void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream)
\r
984 CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );
\r
986 typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, GpuMat& buf, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);
\r
987 static const hist_t hist_callers[] =
\r
989 NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,
\r
991 NppHistogramEvenC1<CV_16U, nppiHistogramEven_16u_C1R, nppiHistogramEvenGetBufferSize_16u_C1R>::hist,
\r
992 NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist
\r
995 hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
\r
998 void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
\r
1001 histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
\r
1004 void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
\r
1006 CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );
\r
1008 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
1009 static const hist_t hist_callers[] =
\r
1011 NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,
\r
1013 NppHistogramEvenC4<CV_16U, nppiHistogramEven_16u_C4R, nppiHistogramEvenGetBufferSize_16u_C4R>::hist,
\r
1014 NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist
\r
1017 hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
\r
1020 void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream)
\r
1023 histRange(src, hist, levels, buf, stream);
\r
1027 void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream)
\r
1029 CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1);
\r
1031 typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, cudaStream_t stream);
\r
1032 static const hist_t hist_callers[] =
\r
1034 NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,
\r
1036 NppHistogramRangeC1<CV_16U, nppiHistogramRange_16u_C1R, nppiHistogramRangeGetBufferSize_16u_C1R>::hist,
\r
1037 NppHistogramRangeC1<CV_16S, nppiHistogramRange_16s_C1R, nppiHistogramRangeGetBufferSize_16s_C1R>::hist,
\r
1039 NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist
\r
1042 hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
\r
1045 void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream)
\r
1048 histRange(src, hist, levels, buf, stream);
\r
1051 void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream)
\r
1053 CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4);
\r
1055 typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, cudaStream_t stream);
\r
1056 static const hist_t hist_callers[] =
\r
1058 NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,
\r
1060 NppHistogramRangeC4<CV_16U, nppiHistogramRange_16u_C4R, nppiHistogramRangeGetBufferSize_16u_C4R>::hist,
\r
1061 NppHistogramRangeC4<CV_16S, nppiHistogramRange_16s_C4R, nppiHistogramRangeGetBufferSize_16s_C4R>::hist,
\r
1063 NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist
\r
1066 hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
\r
1069 namespace cv { namespace gpu { namespace histograms
\r
1071 void histogram256_gpu(DevMem2D src, int* hist, unsigned int* buf, cudaStream_t stream);
\r
1073 const int PARTIAL_HISTOGRAM256_COUNT = 240;
\r
1074 const int HISTOGRAM256_BIN_COUNT = 256;
\r
1077 void cv::gpu::calcHist(const GpuMat& src, GpuMat& hist, Stream& stream)
\r
1080 calcHist(src, hist, buf, stream);
\r
1083 void cv::gpu::calcHist(const GpuMat& src, GpuMat& hist, GpuMat& buf, Stream& stream)
\r
1085 using namespace cv::gpu::histograms;
\r
1087 CV_Assert(src.type() == CV_8UC1);
\r
1089 hist.create(1, 256, CV_32SC1);
\r
1091 ensureSizeIsEnough(1, PARTIAL_HISTOGRAM256_COUNT * HISTOGRAM256_BIN_COUNT, CV_32SC1, buf);
\r
1093 histogram256_gpu(src, hist.ptr<int>(), buf.ptr<unsigned int>(), StreamAccessor::getStream(stream));
\r
1096 void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, Stream& stream)
\r
1100 equalizeHist(src, dst, hist, buf, stream);
\r
1103 void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, Stream& stream)
\r
1106 equalizeHist(src, dst, hist, buf, stream);
\r
1109 namespace cv { namespace gpu { namespace histograms
\r
1111 void equalizeHist_gpu(DevMem2D src, DevMem2D dst, const int* lut, cudaStream_t stream);
\r
1114 void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, GpuMat& buf, Stream& s)
\r
1116 using namespace cv::gpu::histograms;
\r
1118 CV_Assert(src.type() == CV_8UC1);
\r
1120 dst.create(src.size(), src.type());
\r
1123 nppSafeCall( nppsIntegralGetBufferSize_32s(256, &intBufSize) );
\r
1125 int bufSize = static_cast<int>(std::max(256 * 240 * sizeof(int), intBufSize + 256 * sizeof(int)));
\r
1127 ensureSizeIsEnough(1, bufSize, CV_8UC1, buf);
\r
1129 GpuMat histBuf(1, 256 * 240, CV_32SC1, buf.ptr());
\r
1130 GpuMat intBuf(1, intBufSize, CV_8UC1, buf.ptr());
\r
1131 GpuMat lut(1, 256, CV_32S, buf.ptr() + intBufSize);
\r
1133 calcHist(src, hist, histBuf, s);
\r
1135 cudaStream_t stream = StreamAccessor::getStream(s);
\r
1137 NppStreamHandler h(stream);
\r
1139 nppSafeCall( nppsIntegral_32s(hist.ptr<Npp32s>(), lut.ptr<Npp32s>(), 256, intBuf.ptr<Npp8u>()) );
\r
1142 cudaSafeCall( cudaDeviceSynchronize() );
\r
1144 equalizeHist_gpu(src, dst, lut.ptr<int>(), stream);
\r
1147 ////////////////////////////////////////////////////////////////////////
\r
1148 // cornerHarris & minEgenVal
\r
1150 namespace cv { namespace gpu { namespace imgproc {
\r
1152 void extractCovData_caller(const DevMem2Df Dx, const DevMem2Df Dy, PtrStepf dst);
\r
1153 void cornerHarris_caller(const int block_size, const float k, const DevMem2D Dx, const DevMem2D Dy, DevMem2D dst, int border_type);
\r
1154 void cornerMinEigenVal_caller(const int block_size, const DevMem2D Dx, const DevMem2D Dy, DevMem2D dst, int border_type);
\r
1160 template <typename T>
\r
1161 void extractCovData(const GpuMat& src, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
\r
1163 double scale = (double)(1 << ((ksize > 0 ? ksize : 3) - 1)) * blockSize;
\r
1166 if (src.depth() == CV_8U)
\r
1170 Dx.create(src.size(), CV_32F);
\r
1171 Dy.create(src.size(), CV_32F);
\r
1175 Sobel(src, Dx, CV_32F, 1, 0, ksize, scale, borderType);
\r
1176 Sobel(src, Dy, CV_32F, 0, 1, ksize, scale, borderType);
\r
1180 Scharr(src, Dx, CV_32F, 1, 0, scale, borderType);
\r
1181 Scharr(src, Dy, CV_32F, 0, 1, scale, borderType);
\r
1185 void extractCovData(const GpuMat& src, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
\r
1187 switch (src.type())
\r
1190 extractCovData<unsigned char>(src, Dx, Dy, blockSize, ksize, borderType);
\r
1193 extractCovData<float>(src, Dx, Dy, blockSize, ksize, borderType);
\r
1196 CV_Error(CV_StsBadArg, "extractCovData: unsupported type of the source matrix");
\r
1200 } // Anonymous namespace
\r
1203 bool cv::gpu::tryConvertToGpuBorderType(int cpuBorderType, int& gpuBorderType)
\r
1205 if (cpuBorderType == cv::BORDER_REFLECT101)
\r
1207 gpuBorderType = cv::gpu::BORDER_REFLECT101_GPU;
\r
1211 if (cpuBorderType == cv::BORDER_REPLICATE)
\r
1213 gpuBorderType = cv::gpu::BORDER_REPLICATE_GPU;
\r
1217 if (cpuBorderType == cv::BORDER_CONSTANT)
\r
1219 gpuBorderType = cv::gpu::BORDER_CONSTANT_GPU;
\r
1226 void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType)
\r
1229 cornerHarris(src, dst, Dx, Dy, blockSize, ksize, k, borderType);
\r
1232 void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType)
\r
1234 CV_Assert(borderType == cv::BORDER_REFLECT101 ||
\r
1235 borderType == cv::BORDER_REPLICATE);
\r
1237 int gpuBorderType;
\r
1238 CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
\r
1240 extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
\r
1241 dst.create(src.size(), CV_32F);
\r
1242 imgproc::cornerHarris_caller(blockSize, (float)k, Dx, Dy, dst, gpuBorderType);
\r
1245 void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType)
\r
1248 cornerMinEigenVal(src, dst, Dx, Dy, blockSize, ksize, borderType);
\r
1251 void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
\r
1253 CV_Assert(borderType == cv::BORDER_REFLECT101 ||
\r
1254 borderType == cv::BORDER_REPLICATE);
\r
1256 int gpuBorderType;
\r
1257 CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
\r
1259 extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
\r
1260 dst.create(src.size(), CV_32F);
\r
1261 imgproc::cornerMinEigenVal_caller(blockSize, Dx, Dy, dst, gpuBorderType);
\r
1264 //////////////////////////////////////////////////////////////////////////////
\r
1267 namespace cv { namespace gpu { namespace imgproc
\r
1269 void mulSpectrums(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
\r
1270 DevMem2D_<cufftComplex> c);
\r
1272 void mulSpectrums_CONJ(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
\r
1273 DevMem2D_<cufftComplex> c);
\r
1277 void cv::gpu::mulSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c,
\r
1278 int flags, bool conjB)
\r
1280 typedef void (*Caller)(const PtrStep_<cufftComplex>, const PtrStep_<cufftComplex>,
\r
1281 DevMem2D_<cufftComplex>);
\r
1282 static Caller callers[] = { imgproc::mulSpectrums,
\r
1283 imgproc::mulSpectrums_CONJ };
\r
1285 CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
\r
1286 CV_Assert(a.size() == b.size());
\r
1288 c.create(a.size(), CV_32FC2);
\r
1290 Caller caller = callers[(int)conjB];
\r
1294 //////////////////////////////////////////////////////////////////////////////
\r
1295 // mulAndScaleSpectrums
\r
1297 namespace cv { namespace gpu { namespace imgproc
\r
1299 void mulAndScaleSpectrums(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
\r
1300 float scale, DevMem2D_<cufftComplex> c);
\r
1302 void mulAndScaleSpectrums_CONJ(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
\r
1303 float scale, DevMem2D_<cufftComplex> c);
\r
1307 void cv::gpu::mulAndScaleSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c,
\r
1308 int flags, float scale, bool conjB)
\r
1310 typedef void (*Caller)(const PtrStep_<cufftComplex>, const PtrStep_<cufftComplex>,
\r
1311 float scale, DevMem2D_<cufftComplex>);
\r
1312 static Caller callers[] = { imgproc::mulAndScaleSpectrums,
\r
1313 imgproc::mulAndScaleSpectrums_CONJ };
\r
1315 CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
\r
1316 CV_Assert(a.size() == b.size());
\r
1318 c.create(a.size(), CV_32FC2);
\r
1320 Caller caller = callers[(int)conjB];
\r
1321 caller(a, b, scale, c);
\r
1324 //////////////////////////////////////////////////////////////////////////////
\r
1327 void cv::gpu::dft(const GpuMat& src, GpuMat& dst, Size dft_size, int flags)
\r
1329 CV_Assert(src.type() == CV_32F || src.type() == CV_32FC2);
\r
1331 // We don't support unpacked output (in the case of real input)
\r
1332 CV_Assert(!(flags & DFT_COMPLEX_OUTPUT));
\r
1334 bool is_1d_input = (dft_size.height == 1) || (dft_size.width == 1);
\r
1335 int is_row_dft = flags & DFT_ROWS;
\r
1336 int is_scaled_dft = flags & DFT_SCALE;
\r
1337 int is_inverse = flags & DFT_INVERSE;
\r
1338 bool is_complex_input = src.channels() == 2;
\r
1339 bool is_complex_output = !(flags & DFT_REAL_OUTPUT);
\r
1341 // We don't support real-to-real transform
\r
1342 CV_Assert(is_complex_input || is_complex_output);
\r
1346 // Make sure here we work with the continuous input,
\r
1347 // as CUFFT can't handle gaps
\r
1349 createContinuous(src.rows, src.cols, src.type(), src_data);
\r
1350 if (src_data.data != src.data)
\r
1351 src.copyTo(src_data);
\r
1353 Size dft_size_opt = dft_size;
\r
1354 if (is_1d_input && !is_row_dft)
\r
1356 // If the source matrix is single column handle it as single row
\r
1357 dft_size_opt.width = std::max(dft_size.width, dft_size.height);
\r
1358 dft_size_opt.height = std::min(dft_size.width, dft_size.height);
\r
1361 cufftType dft_type = CUFFT_R2C;
\r
1362 if (is_complex_input)
\r
1363 dft_type = is_complex_output ? CUFFT_C2C : CUFFT_C2R;
\r
1365 CV_Assert(dft_size_opt.width > 1);
\r
1368 if (is_1d_input || is_row_dft)
\r
1369 cufftPlan1d(&plan, dft_size_opt.width, dft_type, dft_size_opt.height);
\r
1371 cufftPlan2d(&plan, dft_size_opt.height, dft_size_opt.width, dft_type);
\r
1373 if (is_complex_input)
\r
1375 if (is_complex_output)
\r
1377 createContinuous(dft_size, CV_32FC2, dst);
\r
1378 cufftSafeCall(cufftExecC2C(
\r
1379 plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftComplex>(),
\r
1380 is_inverse ? CUFFT_INVERSE : CUFFT_FORWARD));
\r
1384 createContinuous(dft_size, CV_32F, dst);
\r
1385 cufftSafeCall(cufftExecC2R(
\r
1386 plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftReal>()));
\r
1391 // We could swap dft_size for efficiency. Here we must reflect it
\r
1392 if (dft_size == dft_size_opt)
\r
1393 createContinuous(Size(dft_size.width / 2 + 1, dft_size.height), CV_32FC2, dst);
\r
1395 createContinuous(Size(dft_size.width, dft_size.height / 2 + 1), CV_32FC2, dst);
\r
1397 cufftSafeCall(cufftExecR2C(
\r
1398 plan, src_data.ptr<cufftReal>(), dst.ptr<cufftComplex>()));
\r
1401 cufftSafeCall(cufftDestroy(plan));
\r
1403 if (is_scaled_dft)
\r
1404 multiply(dst, Scalar::all(1. / dft_size.area()), dst);
\r
1407 //////////////////////////////////////////////////////////////////////////////
\r
1411 void cv::gpu::ConvolveBuf::create(Size image_size, Size templ_size)
\r
1413 result_size = Size(image_size.width - templ_size.width + 1,
\r
1414 image_size.height - templ_size.height + 1);
\r
1415 block_size = estimateBlockSize(result_size, templ_size);
\r
1417 dft_size.width = getOptimalDFTSize(block_size.width + templ_size.width - 1);
\r
1418 dft_size.height = getOptimalDFTSize(block_size.width + templ_size.height - 1);
\r
1419 createContinuous(dft_size, CV_32F, image_block);
\r
1420 createContinuous(dft_size, CV_32F, templ_block);
\r
1421 createContinuous(dft_size, CV_32F, result_data);
\r
1423 spect_len = dft_size.height * (dft_size.width / 2 + 1);
\r
1424 createContinuous(1, spect_len, CV_32FC2, image_spect);
\r
1425 createContinuous(1, spect_len, CV_32FC2, templ_spect);
\r
1426 createContinuous(1, spect_len, CV_32FC2, result_spect);
\r
1428 block_size.width = std::min(dft_size.width - templ_size.width + 1, result_size.width);
\r
1429 block_size.height = std::min(dft_size.height - templ_size.height + 1, result_size.height);
\r
1433 Size cv::gpu::ConvolveBuf::estimateBlockSize(Size result_size, Size templ_size)
\r
1436 Size bsize_min(1024, 1024);
\r
1438 // Check whether we use Fermi generation or newer GPU
\r
1439 if (DeviceInfo().majorVersion() >= 2)
\r
1441 bsize_min.width = 2048;
\r
1442 bsize_min.height = 2048;
\r
1445 Size bsize(std::max(templ_size.width * scale, bsize_min.width),
\r
1446 std::max(templ_size.height * scale, bsize_min.height));
\r
1448 bsize.width = std::min(bsize.width, result_size.width);
\r
1449 bsize.height = std::min(bsize.height, result_size.height);
\r
1454 void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
\r
1458 convolve(image, templ, result, ccorr, buf);
\r
1462 void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
\r
1463 bool ccorr, ConvolveBuf& buf)
\r
1465 StaticAssert<sizeof(float) == sizeof(cufftReal)>::check();
\r
1466 StaticAssert<sizeof(float) * 2 == sizeof(cufftComplex)>::check();
\r
1468 CV_Assert(image.type() == CV_32F);
\r
1469 CV_Assert(templ.type() == CV_32F);
\r
1471 buf.create(image.size(), templ.size());
\r
1472 result.create(buf.result_size, CV_32F);
\r
1474 Size& block_size = buf.block_size;
\r
1475 Size& dft_size = buf.dft_size;
\r
1477 GpuMat& image_block = buf.image_block;
\r
1478 GpuMat& templ_block = buf.templ_block;
\r
1479 GpuMat& result_data = buf.result_data;
\r
1481 GpuMat& image_spect = buf.image_spect;
\r
1482 GpuMat& templ_spect = buf.templ_spect;
\r
1483 GpuMat& result_spect = buf.result_spect;
\r
1485 cufftHandle planR2C, planC2R;
\r
1486 cufftSafeCall(cufftPlan2d(&planC2R, dft_size.height, dft_size.width, CUFFT_C2R));
\r
1487 cufftSafeCall(cufftPlan2d(&planR2C, dft_size.height, dft_size.width, CUFFT_R2C));
\r
1489 GpuMat templ_roi(templ.size(), CV_32F, templ.data, templ.step);
\r
1490 copyMakeBorder(templ_roi, templ_block, 0, templ_block.rows - templ_roi.rows, 0,
\r
1491 templ_block.cols - templ_roi.cols, 0);
\r
1493 cufftSafeCall(cufftExecR2C(planR2C, templ_block.ptr<cufftReal>(),
\r
1494 templ_spect.ptr<cufftComplex>()));
\r
1496 // Process all blocks of the result matrix
\r
1497 for (int y = 0; y < result.rows; y += block_size.height)
\r
1499 for (int x = 0; x < result.cols; x += block_size.width)
\r
1501 Size image_roi_size(std::min(x + dft_size.width, image.cols) - x,
\r
1502 std::min(y + dft_size.height, image.rows) - y);
\r
1503 GpuMat image_roi(image_roi_size, CV_32F, (void*)(image.ptr<float>(y) + x),
\r
1505 copyMakeBorder(image_roi, image_block, 0, image_block.rows - image_roi.rows,
\r
1506 0, image_block.cols - image_roi.cols, 0);
\r
1508 cufftSafeCall(cufftExecR2C(planR2C, image_block.ptr<cufftReal>(),
\r
1509 image_spect.ptr<cufftComplex>()));
\r
1510 mulAndScaleSpectrums(image_spect, templ_spect, result_spect, 0,
\r
1511 1.f / dft_size.area(), ccorr);
\r
1512 cufftSafeCall(cufftExecC2R(planC2R, result_spect.ptr<cufftComplex>(),
\r
1513 result_data.ptr<cufftReal>()));
\r
1515 Size result_roi_size(std::min(x + block_size.width, result.cols) - x,
\r
1516 std::min(y + block_size.height, result.rows) - y);
\r
1517 GpuMat result_roi(result_roi_size, result.type(),
\r
1518 (void*)(result.ptr<float>(y) + x), result.step);
\r
1519 GpuMat result_block(result_roi_size, result_data.type(),
\r
1520 result_data.ptr(), result_data.step);
\r
1521 result_block.copyTo(result_roi);
\r
1525 cufftSafeCall(cufftDestroy(planR2C));
\r
1526 cufftSafeCall(cufftDestroy(planC2R));
\r
1530 ////////////////////////////////////////////////////////////////////
\r
1533 namespace cv { namespace gpu { namespace imgproc
\r
1535 template <typename T, int cn>
\r
1536 void downsampleCaller(const DevMem2D src, DevMem2D dst, cudaStream_t stream);
\r
1540 void cv::gpu::downsample(const GpuMat& src, GpuMat& dst, Stream& stream)
\r
1542 CV_Assert(src.depth() < CV_64F && src.channels() <= 4);
\r
1544 typedef void (*Caller)(const DevMem2D, DevMem2D, cudaStream_t stream);
\r
1545 static const Caller callers[6][4] =
\r
1546 {{imgproc::downsampleCaller<uchar,1>, imgproc::downsampleCaller<uchar,2>,
\r
1547 imgproc::downsampleCaller<uchar,3>, imgproc::downsampleCaller<uchar,4>},
\r
1548 {0,0,0,0}, {0,0,0,0},
\r
1549 {imgproc::downsampleCaller<short,1>, imgproc::downsampleCaller<short,2>,
\r
1550 imgproc::downsampleCaller<short,3>, imgproc::downsampleCaller<short,4>},
\r
1552 {imgproc::downsampleCaller<float,1>, imgproc::downsampleCaller<float,2>,
\r
1553 imgproc::downsampleCaller<float,3>, imgproc::downsampleCaller<float,4>}};
\r
1555 Caller caller = callers[src.depth()][src.channels()-1];
\r
1557 CV_Error(CV_StsUnsupportedFormat, "bad number of channels");
\r
1559 dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
\r
1560 caller(src, dst.reshape(1), StreamAccessor::getStream(stream));
\r
1564 //////////////////////////////////////////////////////////////////////////////
\r
1567 namespace cv { namespace gpu { namespace imgproc
\r
1569 template <typename T, int cn>
\r
1570 void upsampleCaller(const DevMem2D src, DevMem2D dst, cudaStream_t stream);
\r
1574 void cv::gpu::upsample(const GpuMat& src, GpuMat& dst, Stream& stream)
\r
1576 CV_Assert(src.depth() < CV_64F && src.channels() <= 4);
\r
1578 typedef void (*Caller)(const DevMem2D, DevMem2D, cudaStream_t stream);
\r
1579 static const Caller callers[6][5] =
\r
1580 {{imgproc::upsampleCaller<uchar,1>, imgproc::upsampleCaller<uchar,2>,
\r
1581 imgproc::upsampleCaller<uchar,3>, imgproc::upsampleCaller<uchar,4>},
\r
1582 {0,0,0,0}, {0,0,0,0},
\r
1583 {imgproc::upsampleCaller<short,1>, imgproc::upsampleCaller<short,2>,
\r
1584 imgproc::upsampleCaller<short,3>, imgproc::upsampleCaller<short,4>},
\r
1586 {imgproc::upsampleCaller<float,1>, imgproc::upsampleCaller<float,2>,
\r
1587 imgproc::upsampleCaller<float,3>, imgproc::upsampleCaller<float,4>}};
\r
1589 Caller caller = callers[src.depth()][src.channels()-1];
\r
1591 CV_Error(CV_StsUnsupportedFormat, "bad number of channels");
\r
1593 dst.create(src.rows*2, src.cols*2, src.type());
\r
1594 caller(src, dst.reshape(1), StreamAccessor::getStream(stream));
\r
1598 //////////////////////////////////////////////////////////////////////////////
\r
1601 void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, Stream& stream)
\r
1604 pyrDown(src, dst, buf, stream);
\r
1607 cv::Mat cv::gpu::PyrDownBuf::ker;
\r
1609 void cv::gpu::PyrDownBuf::create(Size image_size, int image_type_)
\r
1611 if (ker.empty() || image_type_ != image_type)
\r
1612 ker = getGaussianKernel(5, 0, std::max(CV_32F, CV_MAT_DEPTH(image_type_)));
\r
1614 ensureSizeIsEnough(image_size.height, image_size.width, image_type_, buf);
\r
1616 if (filter.empty() || image_type_ != image_type)
\r
1618 image_type = image_type_;
\r
1619 filter = createSeparableLinearFilter_GPU(image_type, image_type, ker, ker);
\r
1623 void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, PyrDownBuf& buf, Stream& stream)
\r
1625 buf.create(src.size(), src.type());
\r
1626 buf.filter->apply(src, buf.buf, Rect(0, 0, src.cols, src.rows), stream);
\r
1627 downsample(buf.buf, dst, stream);
\r
1631 //////////////////////////////////////////////////////////////////////////////
\r
1634 void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, Stream& stream)
\r
1637 pyrUp(src, dst, buf, stream);
\r
1640 cv::Mat cv::gpu::PyrUpBuf::ker;
\r
1642 void cv::gpu::PyrUpBuf::create(Size image_size, int image_type_)
\r
1644 if (ker.empty() || image_type_ != image_type)
\r
1645 ker = getGaussianKernel(5, 0, std::max(CV_32F, CV_MAT_DEPTH(image_type_))) * 2;
\r
1647 ensureSizeIsEnough(image_size.height * 2, image_size.width * 2, image_type_, buf);
\r
1649 if (filter.empty() || image_type_ != image_type)
\r
1651 image_type = image_type_;
\r
1652 filter = createSeparableLinearFilter_GPU(image_type, image_type, ker, ker);
\r
1656 void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, PyrUpBuf& buf, Stream& stream)
\r
1658 buf.create(src.size(), src.type());
\r
1659 upsample(src, buf.buf, stream);
\r
1660 buf.filter->apply(buf.buf, dst, Rect(0, 0, buf.buf.cols, buf.buf.rows), stream);
\r
1664 //////////////////////////////////////////////////////////////////////////////
\r
1667 cv::gpu::CannyBuf::CannyBuf(const GpuMat& dx_, const GpuMat& dy_) : dx(dx_), dy(dy_)
\r
1669 CV_Assert(dx_.type() == CV_32SC1 && dy_.type() == CV_32SC1 && dx_.size() == dy_.size());
\r
1671 create(dx_.size(), -1);
\r
1674 void cv::gpu::CannyBuf::create(const Size& image_size, int apperture_size)
\r
1676 ensureSizeIsEnough(image_size, CV_32SC1, dx);
\r
1677 ensureSizeIsEnough(image_size, CV_32SC1, dy);
\r
1679 if (apperture_size == 3)
\r
1681 ensureSizeIsEnough(image_size, CV_32SC1, dx_buf);
\r
1682 ensureSizeIsEnough(image_size, CV_32SC1, dy_buf);
\r
1684 else if(apperture_size > 0)
\r
1687 filterDX = createDerivFilter_GPU(CV_8UC1, CV_32S, 1, 0, apperture_size, BORDER_REPLICATE);
\r
1689 filterDY = createDerivFilter_GPU(CV_8UC1, CV_32S, 0, 1, apperture_size, BORDER_REPLICATE);
\r
1692 ensureSizeIsEnough(image_size.height + 2, image_size.width + 2, CV_32FC1, edgeBuf);
\r
1694 ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf1);
\r
1695 ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf2);
\r
1698 void cv::gpu::CannyBuf::release()
\r
1704 edgeBuf.release();
\r
1705 trackBuf1.release();
\r
1706 trackBuf2.release();
\r
1709 namespace cv { namespace gpu { namespace canny
\r
1711 void calcSobelRowPass_gpu(PtrStep src, PtrStepi dx_buf, PtrStepi dy_buf, int rows, int cols);
\r
1713 void calcMagnitude_gpu(PtrStepi dx_buf, PtrStepi dy_buf, PtrStepi dx, PtrStepi dy, PtrStepf mag, int rows, int cols, bool L2Grad);
\r
1714 void calcMagnitude_gpu(PtrStepi dx, PtrStepi dy, PtrStepf mag, int rows, int cols, bool L2Grad);
\r
1716 void calcMap_gpu(PtrStepi dx, PtrStepi dy, PtrStepf mag, PtrStepi map, int rows, int cols, float low_thresh, float high_thresh);
\r
1718 void edgesHysteresisLocal_gpu(PtrStepi map, ushort2* st1, int rows, int cols);
\r
1720 void edgesHysteresisGlobal_gpu(PtrStepi map, ushort2* st1, ushort2* st2, int rows, int cols);
\r
1722 void getEdges_gpu(PtrStepi map, PtrStep dst, int rows, int cols);
\r
1727 void CannyCaller(CannyBuf& buf, GpuMat& dst, float low_thresh, float high_thresh)
\r
1729 using namespace cv::gpu::canny;
\r
1731 calcMap_gpu(buf.dx, buf.dy, buf.edgeBuf, buf.edgeBuf, dst.rows, dst.cols, low_thresh, high_thresh);
\r
1733 edgesHysteresisLocal_gpu(buf.edgeBuf, buf.trackBuf1.ptr<ushort2>(), dst.rows, dst.cols);
\r
1735 edgesHysteresisGlobal_gpu(buf.edgeBuf, buf.trackBuf1.ptr<ushort2>(), buf.trackBuf2.ptr<ushort2>(), dst.rows, dst.cols);
\r
1737 getEdges_gpu(buf.edgeBuf, dst, dst.rows, dst.cols);
\r
1741 void cv::gpu::Canny(const GpuMat& src, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
\r
1743 CannyBuf buf(src.size(), apperture_size);
\r
1744 Canny(src, buf, dst, low_thresh, high_thresh, apperture_size, L2gradient);
\r
1747 void cv::gpu::Canny(const GpuMat& src, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
\r
1749 using namespace cv::gpu::canny;
\r
1751 CV_Assert(src.type() == CV_8UC1);
\r
1753 if( low_thresh > high_thresh )
\r
1754 std::swap( low_thresh, high_thresh);
\r
1756 dst.create(src.size(), CV_8U);
\r
1757 dst.setTo(Scalar::all(0));
\r
1759 buf.create(src.size(), apperture_size);
\r
1760 buf.edgeBuf.setTo(Scalar::all(0));
\r
1762 if (apperture_size == 3)
\r
1764 calcSobelRowPass_gpu(src, buf.dx_buf, buf.dy_buf, src.rows, src.cols);
\r
1766 calcMagnitude_gpu(buf.dx_buf, buf.dy_buf, buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
\r
1770 buf.filterDX->apply(src, buf.dx, Rect(0, 0, src.cols, src.rows));
\r
1771 buf.filterDY->apply(src, buf.dy, Rect(0, 0, src.cols, src.rows));
\r
1773 calcMagnitude_gpu(buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
\r
1776 CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
\r
1779 void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
\r
1781 CannyBuf buf(dx, dy);
\r
1782 Canny(dx, dy, buf, dst, low_thresh, high_thresh, L2gradient);
\r
1785 void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
\r
1787 using namespace cv::gpu::canny;
\r
1789 CV_Assert(dx.type() == CV_32SC1 && dy.type() == CV_32SC1 && dx.size() == dy.size());
\r
1791 if( low_thresh > high_thresh )
\r
1792 std::swap( low_thresh, high_thresh);
\r
1794 dst.create(dx.size(), CV_8U);
\r
1795 dst.setTo(Scalar::all(0));
\r
1797 buf.dx = dx; buf.dy = dy;
\r
1798 buf.create(dx.size(), -1);
\r
1799 buf.edgeBuf.setTo(Scalar::all(0));
\r
1801 calcMagnitude_gpu(dx, dy, buf.edgeBuf, dx.rows, dx.cols, L2gradient);
\r
1803 CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
\r
1806 #endif /* !defined (HAVE_CUDA) */
\r