2010-07-23 15:06:33 +08:00
|
|
|
/*M///////////////////////////////////////////////////////////////////////////////////////
|
|
|
|
//
|
|
|
|
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
|
|
|
|
//
|
|
|
|
// By downloading, copying, installing or using the software you agree to this license.
|
|
|
|
// If you do not agree to this license, do not download, install,
|
|
|
|
// copy or use the software.
|
|
|
|
//
|
|
|
|
//
|
|
|
|
// License Agreement
|
|
|
|
// For Open Source Computer Vision Library
|
|
|
|
//
|
|
|
|
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
|
|
|
|
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
|
|
|
|
// Third party copyrights are property of their respective owners.
|
|
|
|
//
|
|
|
|
// Redistribution and use in source and binary forms, with or without modification,
|
|
|
|
// are permitted provided that the following conditions are met:
|
|
|
|
//
|
|
|
|
// * Redistribution's of source code must retain the above copyright notice,
|
|
|
|
// this list of conditions and the following disclaimer.
|
|
|
|
//
|
|
|
|
// * Redistribution's in binary form must reproduce the above copyright notice,
|
|
|
|
// this list of conditions and the following disclaimer in the documentation
|
|
|
|
// and/or other materials provided with the distribution.
|
|
|
|
//
|
|
|
|
// * The name of the copyright holders may not be used to endorse or promote products
|
|
|
|
// derived from this software without specific prior written permission.
|
|
|
|
//
|
|
|
|
// This software is provided by the copyright holders and contributors "as is" and
|
|
|
|
// any express or implied warranties, including, but not limited to, the implied
|
|
|
|
// warranties of merchantability and fitness for a particular purpose are disclaimed.
|
|
|
|
// In no event shall the Intel Corporation or contributors be liable for any direct,
|
|
|
|
// indirect, incidental, special, exemplary, or consequential damages
|
|
|
|
// (including, but not limited to, procurement of substitute goods or services;
|
|
|
|
// loss of use, data, or profits; or business interruption) however caused
|
|
|
|
// and on any theory of liability, whether in contract, strict liability,
|
|
|
|
// or tort (including negligence or otherwise) arising in any way out of
|
|
|
|
// the use of this software, even if advised of the possibility of such damage.
|
|
|
|
//
|
|
|
|
//M*/
|
|
|
|
|
|
|
|
#include "precomp.hpp"
|
|
|
|
|
|
|
|
using namespace cv;
|
|
|
|
using namespace cv::gpu;
|
|
|
|
|
|
|
|
#if !defined (HAVE_CUDA)
|
|
|
|
|
2011-09-05 15:51:00 +08:00
|
|
|
void cv::gpu::remap(const GpuMat&, GpuMat&, const GpuMat&, const GpuMat&, int, int, const Scalar&, Stream&){ throw_nogpu(); }
|
2010-09-15 20:47:59 +08:00
|
|
|
void cv::gpu::meanShiftFiltering(const GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
|
2010-10-11 22:25:30 +08:00
|
|
|
void cv::gpu::meanShiftProc(const GpuMat&, GpuMat&, GpuMat&, int, int, TermCriteria) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::drawColorDisp(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::reprojectImageTo3D(const GpuMat&, GpuMat&, const Mat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::resize(const GpuMat&, GpuMat&, Size, double, double, int, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::copyMakeBorder(const GpuMat&, GpuMat&, int, int, int, int, const Scalar&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::warpAffine(const GpuMat&, GpuMat&, const Mat&, Size, int, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::warpPerspective(const GpuMat&, GpuMat&, const Mat&, Size, int, Stream&) { throw_nogpu(); }
|
2011-07-01 15:07:54 +08:00
|
|
|
void cv::gpu::buildWarpPlaneMaps(Size, Rect, const Mat&, double, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::buildWarpCylindricalMaps(Size, Rect, const Mat&, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::buildWarpSphericalMaps(Size, Rect, const Mat&, double, double, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::rotate(const GpuMat&, GpuMat&, Size, double, double, double, int, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::integral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::integralBuffered(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::integral(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::sqrIntegral(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2010-12-08 23:06:10 +08:00
|
|
|
void cv::gpu::columnSum(const GpuMat&, GpuMat&) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::rectStdDev(const GpuMat&, const GpuMat&, GpuMat&, const Rect&, Stream&) { throw_nogpu(); }
|
2010-10-12 14:37:44 +08:00
|
|
|
void cv::gpu::evenLevels(GpuMat&, int, int, int) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat&, GpuMat&, GpuMat&, int, int, int, Stream&) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat&, GpuMat*, int*, int*, int*, Stream&) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat&, GpuMat*, GpuMat&, int*, int*, int*, Stream&) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, Stream&) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat&, GpuMat&, const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, Stream&) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat&, GpuMat*, const GpuMat*, GpuMat&, Stream&) { throw_nogpu(); }
|
2011-08-01 21:15:37 +08:00
|
|
|
void cv::gpu::calcHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::calcHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2011-08-02 16:33:27 +08:00
|
|
|
void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::equalizeHist(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2010-12-02 17:07:13 +08:00
|
|
|
void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::cornerHarris(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, double, int) { throw_nogpu(); }
|
2010-12-02 17:07:13 +08:00
|
|
|
void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
|
2011-08-08 21:05:50 +08:00
|
|
|
void cv::gpu::cornerMinEigenVal(const GpuMat&, GpuMat&, GpuMat&, GpuMat&, int, int, int) { throw_nogpu(); }
|
2010-12-22 21:46:06 +08:00
|
|
|
void cv::gpu::mulSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, bool) { throw_nogpu(); }
|
|
|
|
void cv::gpu::mulAndScaleSpectrums(const GpuMat&, const GpuMat&, GpuMat&, int, float, bool) { throw_nogpu(); }
|
2010-12-27 15:35:41 +08:00
|
|
|
void cv::gpu::dft(const GpuMat&, GpuMat&, Size, int) { throw_nogpu(); }
|
2010-12-27 18:18:42 +08:00
|
|
|
void cv::gpu::ConvolveBuf::create(Size, Size) { throw_nogpu(); }
|
2010-12-22 21:46:06 +08:00
|
|
|
void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool) { throw_nogpu(); }
|
2010-12-27 18:18:42 +08:00
|
|
|
void cv::gpu::convolve(const GpuMat&, const GpuMat&, GpuMat&, bool, ConvolveBuf&) { throw_nogpu(); }
|
2011-08-01 16:15:31 +08:00
|
|
|
void cv::gpu::downsample(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::upsample(const GpuMat&, GpuMat&, Stream&) { throw_nogpu(); }
|
2011-08-31 19:42:54 +08:00
|
|
|
void cv::gpu::pyrDown(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::pyrUp(const GpuMat&, GpuMat&, int, Stream&) { throw_nogpu(); }
|
2011-08-08 16:53:55 +08:00
|
|
|
void cv::gpu::Canny(const GpuMat&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
|
|
|
|
void cv::gpu::Canny(const GpuMat&, CannyBuf&, GpuMat&, double, double, int, bool) { throw_nogpu(); }
|
|
|
|
void cv::gpu::Canny(const GpuMat&, const GpuMat&, GpuMat&, double, double, bool) { throw_nogpu(); }
|
|
|
|
void cv::gpu::Canny(const GpuMat&, const GpuMat&, CannyBuf&, GpuMat&, double, double, bool) { throw_nogpu(); }
|
|
|
|
cv::gpu::CannyBuf::CannyBuf(const GpuMat&, const GpuMat&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::CannyBuf::create(const Size&, int) { throw_nogpu(); }
|
|
|
|
void cv::gpu::CannyBuf::release() { throw_nogpu(); }
|
2010-07-23 15:06:33 +08:00
|
|
|
|
|
|
|
#else /* !defined (HAVE_CUDA) */
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// remap
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
2010-10-26 17:37:24 +08:00
|
|
|
{
|
2011-08-31 19:42:54 +08:00
|
|
|
template <typename T> void remap_gpu(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D& dst,
|
2011-09-05 15:51:00 +08:00
|
|
|
int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
|
2011-08-31 19:42:54 +08:00
|
|
|
}}}
|
2010-10-13 17:42:29 +08:00
|
|
|
|
2011-09-05 15:51:00 +08:00
|
|
|
void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode, const Scalar& borderValue, Stream& stream)
|
2011-08-31 19:42:54 +08:00
|
|
|
{
|
|
|
|
using namespace cv::gpu::imgproc;
|
2010-10-13 17:42:29 +08:00
|
|
|
|
2011-09-05 15:51:00 +08:00
|
|
|
typedef void (*caller_t)(const DevMem2D& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);
|
2011-08-31 19:42:54 +08:00
|
|
|
static const caller_t callers[6][4] =
|
|
|
|
{
|
|
|
|
{remap_gpu<uchar>, remap_gpu<uchar2>, remap_gpu<uchar3>, remap_gpu<uchar4>},
|
|
|
|
{remap_gpu<schar>, remap_gpu<char2>, remap_gpu<char3>, remap_gpu<char4>},
|
|
|
|
{remap_gpu<ushort>, remap_gpu<ushort2>, remap_gpu<ushort3>, remap_gpu<ushort4>},
|
|
|
|
{remap_gpu<short>, remap_gpu<short2>, remap_gpu<short3>, remap_gpu<short4>},
|
|
|
|
{remap_gpu<int>, remap_gpu<int2>, remap_gpu<int3>, remap_gpu<int4>},
|
|
|
|
{remap_gpu<float>, remap_gpu<float2>, remap_gpu<float3>, remap_gpu<float4>}
|
|
|
|
};
|
2010-07-23 15:06:33 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
|
|
|
|
CV_Assert(xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size());
|
2010-08-30 21:41:26 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR);
|
2010-08-26 17:19:47 +08:00
|
|
|
|
2011-08-31 21:21:14 +08:00
|
|
|
CV_Assert(borderMode == BORDER_REFLECT101 || borderMode == BORDER_REPLICATE || borderMode == BORDER_CONSTANT || borderMode == BORDER_REFLECT || borderMode == BORDER_WRAP);
|
2011-08-31 19:42:54 +08:00
|
|
|
int gpuBorderType;
|
|
|
|
CV_Assert(tryConvertToGpuBorderType(borderMode, gpuBorderType));
|
2010-07-23 15:06:33 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
dst.create(xmap.size(), src.type());
|
2011-09-05 15:51:00 +08:00
|
|
|
|
|
|
|
Scalar_<float> borderValueFloat;
|
|
|
|
borderValueFloat = borderValue;
|
2010-08-26 17:19:47 +08:00
|
|
|
|
2011-09-05 15:51:00 +08:00
|
|
|
callers[src.depth()][src.channels() - 1](src, xmap, ymap, dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream));
|
2010-07-23 15:06:33 +08:00
|
|
|
}
|
|
|
|
|
2010-08-30 21:41:26 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// meanShiftFiltering_GPU
|
2010-08-07 01:02:06 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
extern "C" void meanShiftFiltering_gpu(const DevMem2D& src, DevMem2D dst, int sp, int sr, int maxIter, float eps);
|
|
|
|
}}}
|
|
|
|
|
2010-09-06 22:27:23 +08:00
|
|
|
void cv::gpu::meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-08-07 01:02:06 +08:00
|
|
|
if( src.empty() )
|
|
|
|
CV_Error( CV_StsBadArg, "The input image is empty" );
|
|
|
|
|
|
|
|
if( src.depth() != CV_8U || src.channels() != 4 )
|
|
|
|
CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
|
|
|
|
|
2010-09-03 22:32:12 +08:00
|
|
|
dst.create( src.size(), CV_8UC4 );
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-07 01:02:06 +08:00
|
|
|
if( !(criteria.type & TermCriteria::MAX_ITER) )
|
|
|
|
criteria.maxCount = 5;
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-07 01:02:06 +08:00
|
|
|
int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-09-03 22:32:12 +08:00
|
|
|
float eps;
|
2010-08-07 01:02:06 +08:00
|
|
|
if( !(criteria.type & TermCriteria::EPS) )
|
|
|
|
eps = 1.f;
|
2010-11-08 17:55:10 +08:00
|
|
|
eps = (float)std::max(criteria.epsilon, 0.0);
|
2010-09-03 22:32:12 +08:00
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
imgproc::meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps);
|
2010-08-07 01:02:06 +08:00
|
|
|
}
|
|
|
|
|
2010-08-30 21:41:26 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
2010-10-11 22:25:30 +08:00
|
|
|
// meanShiftProc_GPU
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
extern "C" void meanShiftProc_gpu(const DevMem2D& src, DevMem2D dstr, DevMem2D dstsp, int sp, int sr, int maxIter, float eps);
|
|
|
|
}}}
|
|
|
|
|
2010-10-11 22:25:30 +08:00
|
|
|
void cv::gpu::meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-11 22:25:30 +08:00
|
|
|
if( src.empty() )
|
|
|
|
CV_Error( CV_StsBadArg, "The input image is empty" );
|
|
|
|
|
|
|
|
if( src.depth() != CV_8U || src.channels() != 4 )
|
|
|
|
CV_Error( CV_StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
|
|
|
|
|
|
|
|
dstr.create( src.size(), CV_8UC4 );
|
|
|
|
dstsp.create( src.size(), CV_16SC2 );
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-10-11 22:25:30 +08:00
|
|
|
if( !(criteria.type & TermCriteria::MAX_ITER) )
|
|
|
|
criteria.maxCount = 5;
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-10-11 22:25:30 +08:00
|
|
|
int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-10-11 22:25:30 +08:00
|
|
|
float eps;
|
|
|
|
if( !(criteria.type & TermCriteria::EPS) )
|
|
|
|
eps = 1.f;
|
2010-11-08 17:55:10 +08:00
|
|
|
eps = (float)std::max(criteria.epsilon, 0.0);
|
2010-10-11 22:25:30 +08:00
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
imgproc::meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps);
|
2010-10-11 22:25:30 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
////////////////////////////////////////////////////////////////////////
|
2010-08-30 21:41:26 +08:00
|
|
|
// drawColorDisp
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void drawColorDisp_gpu(const DevMem2D& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);
|
|
|
|
void drawColorDisp_gpu(const DevMem2D_<short>& src, const DevMem2D& dst, int ndisp, const cudaStream_t& stream);
|
|
|
|
}}}
|
|
|
|
|
2010-08-20 14:47:11 +08:00
|
|
|
namespace
|
|
|
|
{
|
|
|
|
template <typename T>
|
2010-08-25 14:30:11 +08:00
|
|
|
void drawColorDisp_caller(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2011-08-08 21:05:50 +08:00
|
|
|
dst.create(src.size(), CV_8UC4);
|
2010-08-25 14:30:11 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
imgproc::drawColorDisp_gpu((DevMem2D_<T>)src, dst, ndisp, stream);
|
2010-08-20 14:47:11 +08:00
|
|
|
}
|
2010-08-25 14:30:11 +08:00
|
|
|
|
|
|
|
typedef void (*drawColorDisp_caller_t)(const GpuMat& src, GpuMat& dst, int ndisp, const cudaStream_t& stream);
|
|
|
|
|
|
|
|
const drawColorDisp_caller_t drawColorDisp_callers[] = {drawColorDisp_caller<unsigned char>, 0, 0, drawColorDisp_caller<short>, 0, 0, 0, 0};
|
2010-08-20 14:47:11 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::drawColorDisp(const GpuMat& src, GpuMat& dst, int ndisp, Stream& stream)
|
2010-08-25 14:30:11 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8U || src.type() == CV_16S);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-25 14:30:11 +08:00
|
|
|
drawColorDisp_callers[src.type()](src, dst, ndisp, StreamAccessor::getStream(stream));
|
2010-08-20 14:47:11 +08:00
|
|
|
}
|
2010-08-07 01:02:06 +08:00
|
|
|
|
2010-08-30 21:41:26 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// reprojectImageTo3D
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void reprojectImageTo3D_gpu(const DevMem2D& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
|
|
|
|
void reprojectImageTo3D_gpu(const DevMem2D_<short>& disp, const DevMem2Df& xyzw, const float* q, const cudaStream_t& stream);
|
|
|
|
}}}
|
|
|
|
|
2010-08-23 22:19:22 +08:00
|
|
|
namespace
|
|
|
|
{
|
|
|
|
template <typename T>
|
2010-08-25 14:30:11 +08:00
|
|
|
void reprojectImageTo3D_caller(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-08-25 14:30:11 +08:00
|
|
|
xyzw.create(disp.rows, disp.cols, CV_32FC4);
|
2010-10-20 16:50:14 +08:00
|
|
|
imgproc::reprojectImageTo3D_gpu((DevMem2D_<T>)disp, xyzw, Q.ptr<float>(), stream);
|
2010-08-23 22:19:22 +08:00
|
|
|
}
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-25 14:30:11 +08:00
|
|
|
typedef void (*reprojectImageTo3D_caller_t)(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, const cudaStream_t& stream);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-25 14:30:11 +08:00
|
|
|
const reprojectImageTo3D_caller_t reprojectImageTo3D_callers[] = {reprojectImageTo3D_caller<unsigned char>, 0, 0, reprojectImageTo3D_caller<short>, 0, 0, 0, 0};
|
2010-08-23 22:19:22 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::reprojectImageTo3D(const GpuMat& disp, GpuMat& xyzw, const Mat& Q, Stream& stream)
|
2010-08-25 14:30:11 +08:00
|
|
|
{
|
|
|
|
CV_Assert((disp.type() == CV_8U || disp.type() == CV_16S) && Q.type() == CV_32F && Q.rows == 4 && Q.cols == 4);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-25 14:30:11 +08:00
|
|
|
reprojectImageTo3D_callers[disp.type()](disp, xyzw, Q, StreamAccessor::getStream(stream));
|
2010-08-23 22:19:22 +08:00
|
|
|
}
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// resize
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::resize(const GpuMat& src, GpuMat& dst, Size dsize, double fx, double fy, int interpolation, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-09-27 20:44:57 +08:00
|
|
|
static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR/*, NPPI_INTER_CUBIC, 0, NPPI_INTER_LANCZOS*/};
|
2010-09-22 18:58:01 +08:00
|
|
|
|
|
|
|
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
|
2010-09-27 20:44:57 +08:00
|
|
|
CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR/* || interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4*/);
|
2010-09-22 18:58:01 +08:00
|
|
|
|
|
|
|
CV_Assert( src.size().area() > 0 );
|
|
|
|
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 = (double)dsize.width / src.cols;
|
|
|
|
fy = (double)dsize.height / src.rows;
|
|
|
|
}
|
|
|
|
|
|
|
|
dst.create(dsize, src.type());
|
|
|
|
|
|
|
|
NppiSize srcsz;
|
|
|
|
srcsz.width = src.cols;
|
|
|
|
srcsz.height = src.rows;
|
|
|
|
NppiRect srcrect;
|
|
|
|
srcrect.x = srcrect.y = 0;
|
|
|
|
srcrect.width = src.cols;
|
|
|
|
srcrect.height = src.rows;
|
|
|
|
NppiSize dstsz;
|
|
|
|
dstsz.width = dst.cols;
|
|
|
|
dstsz.height = dst.rows;
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
if (src.type() == CV_8UC1)
|
|
|
|
{
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiResize_8u_C1R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcrect,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, fx, fy, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiResize_8u_C4R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcrect,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, fx, fy, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// copyMakeBorder
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::copyMakeBorder(const GpuMat& src, GpuMat& dst, int top, int bottom, int left, int right, const Scalar& value, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-12-20 20:49:40 +08:00
|
|
|
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4 || src.type() == CV_32SC1 || src.type() == CV_32FC1);
|
2010-09-22 18:58:01 +08:00
|
|
|
|
|
|
|
dst.create(src.rows + top + bottom, src.cols + left + right, src.type());
|
|
|
|
|
2010-12-22 15:30:21 +08:00
|
|
|
NppiSize srcsz;
|
|
|
|
srcsz.width = src.cols;
|
|
|
|
srcsz.height = src.rows;
|
2010-09-22 18:58:01 +08:00
|
|
|
NppiSize dstsz;
|
2010-12-22 15:30:21 +08:00
|
|
|
dstsz.width = dst.cols;
|
|
|
|
dstsz.height = dst.rows;
|
2010-09-22 18:58:01 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
switch (src.type())
|
|
|
|
{
|
|
|
|
case CV_8UC1:
|
2010-12-20 20:49:40 +08:00
|
|
|
{
|
2010-09-22 18:58:01 +08:00
|
|
|
Npp8u nVal = static_cast<Npp8u>(value[0]);
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiCopyConstBorder_8u_C1R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
2010-12-20 20:49:40 +08:00
|
|
|
}
|
2010-09-22 18:58:01 +08:00
|
|
|
case CV_8UC4:
|
2010-12-20 20:49:40 +08:00
|
|
|
{
|
2010-09-22 18:58:01 +08:00
|
|
|
Npp8u nVal[] = {static_cast<Npp8u>(value[0]), static_cast<Npp8u>(value[1]), static_cast<Npp8u>(value[2]), static_cast<Npp8u>(value[3])};
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiCopyConstBorder_8u_C4R(src.ptr<Npp8u>(), static_cast<int>(src.step), srcsz,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
2010-12-20 20:49:40 +08:00
|
|
|
}
|
2010-09-22 18:58:01 +08:00
|
|
|
case CV_32SC1:
|
2010-12-20 20:49:40 +08:00
|
|
|
{
|
2010-09-22 18:58:01 +08:00
|
|
|
Npp32s nVal = static_cast<Npp32s>(value[0]);
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
|
|
|
|
dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
2010-12-20 20:49:40 +08:00
|
|
|
}
|
|
|
|
case CV_32FC1:
|
|
|
|
{
|
2010-12-22 15:30:21 +08:00
|
|
|
Npp32f val = static_cast<Npp32f>(value[0]);
|
2010-12-20 20:49:40 +08:00
|
|
|
Npp32s nVal = *(reinterpret_cast<Npp32s*>(&val));
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiCopyConstBorder_32s_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), srcsz,
|
|
|
|
dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstsz, top, left, nVal) );
|
2010-12-20 20:49:40 +08:00
|
|
|
break;
|
|
|
|
}
|
2010-09-22 18:58:01 +08:00
|
|
|
default:
|
|
|
|
CV_Assert(!"Unsupported source type");
|
|
|
|
}
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// warp
|
|
|
|
|
|
|
|
namespace
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
|
|
|
typedef NppStatus (*npp_warp_8u_t)(const Npp8u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp8u* pDst,
|
|
|
|
int dstStep, NppiRect dstRoi, const double coeffs[][3],
|
2010-09-22 18:58:01 +08:00
|
|
|
int interpolation);
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*npp_warp_16u_t)(const Npp16u* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp16u* pDst,
|
|
|
|
int dstStep, NppiRect dstRoi, const double coeffs[][3],
|
2010-09-22 18:58:01 +08:00
|
|
|
int interpolation);
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*npp_warp_32s_t)(const Npp32s* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32s* pDst,
|
|
|
|
int dstStep, NppiRect dstRoi, const double coeffs[][3],
|
2010-09-22 18:58:01 +08:00
|
|
|
int interpolation);
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*npp_warp_32f_t)(const Npp32f* pSrc, NppiSize srcSize, int srcStep, NppiRect srcRoi, Npp32f* pDst,
|
|
|
|
int dstStep, NppiRect dstRoi, const double coeffs[][3],
|
2010-09-22 18:58:01 +08:00
|
|
|
int interpolation);
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
void nppWarpCaller(const GpuMat& src, GpuMat& dst, double coeffs[][3], const Size& dsize, int flags,
|
|
|
|
npp_warp_8u_t npp_warp_8u[][2], npp_warp_16u_t npp_warp_16u[][2],
|
2011-05-31 16:31:10 +08:00
|
|
|
npp_warp_32s_t npp_warp_32s[][2], npp_warp_32f_t npp_warp_32f[][2], cudaStream_t stream)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
|
|
|
static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
int interpolation = flags & INTER_MAX;
|
|
|
|
|
|
|
|
CV_Assert((src.depth() == CV_8U || src.depth() == CV_16U || src.depth() == CV_32S || src.depth() == CV_32F) && src.channels() != 2);
|
|
|
|
CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
|
|
|
|
|
|
|
|
dst.create(dsize, src.type());
|
|
|
|
|
|
|
|
NppiSize srcsz;
|
|
|
|
srcsz.height = src.rows;
|
|
|
|
srcsz.width = src.cols;
|
|
|
|
NppiRect srcroi;
|
|
|
|
srcroi.x = srcroi.y = 0;
|
|
|
|
srcroi.height = src.rows;
|
|
|
|
srcroi.width = src.cols;
|
|
|
|
NppiRect dstroi;
|
|
|
|
dstroi.x = dstroi.y = 0;
|
|
|
|
dstroi.height = dst.rows;
|
|
|
|
dstroi.width = dst.cols;
|
|
|
|
|
|
|
|
int warpInd = (flags & WARP_INVERSE_MAP) >> 4;
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
switch (src.depth())
|
|
|
|
{
|
|
|
|
case CV_8U:
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( npp_warp_8u[src.channels()][warpInd](src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
|
|
|
case CV_16U:
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( npp_warp_16u[src.channels()][warpInd](src.ptr<Npp16u>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp16u>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
|
|
|
case CV_32S:
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( npp_warp_32s[src.channels()][warpInd](src.ptr<Npp32s>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp32s>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
|
|
|
case CV_32F:
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( npp_warp_32f[src.channels()][warpInd](src.ptr<Npp32f>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp32f>(), static_cast<int>(dst.step), dstroi, coeffs, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
CV_Assert(!"Unsupported source type");
|
|
|
|
}
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::warpAffine(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_8u_t npp_warpAffine_8u[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_8u_C1R, nppiWarpAffineBack_8u_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_8u_C3R, nppiWarpAffineBack_8u_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpAffine_8u_C4R, nppiWarpAffineBack_8u_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_16u_t npp_warpAffine_16u[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_16u_C1R, nppiWarpAffineBack_16u_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_16u_C3R, nppiWarpAffineBack_16u_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpAffine_16u_C4R, nppiWarpAffineBack_16u_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_32s_t npp_warpAffine_32s[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_32s_C1R, nppiWarpAffineBack_32s_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_32s_C3R, nppiWarpAffineBack_32s_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpAffine_32s_C4R, nppiWarpAffineBack_32s_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_32f_t npp_warpAffine_32f[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_32f_C1R, nppiWarpAffineBack_32f_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpAffine_32f_C3R, nppiWarpAffineBack_32f_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpAffine_32f_C4R, nppiWarpAffineBack_32f_C4R}
|
|
|
|
};
|
|
|
|
|
|
|
|
CV_Assert(M.rows == 2 && M.cols == 3);
|
|
|
|
|
|
|
|
double coeffs[2][3];
|
|
|
|
Mat coeffsMat(2, 3, CV_64F, (void*)coeffs);
|
|
|
|
M.convertTo(coeffsMat, coeffsMat.type());
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpAffine_8u, npp_warpAffine_16u, npp_warpAffine_32s, npp_warpAffine_32f, StreamAccessor::getStream(s));
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::warpPerspective(const GpuMat& src, GpuMat& dst, const Mat& M, Size dsize, int flags, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_8u_t npp_warpPerspective_8u[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_8u_C1R, nppiWarpPerspectiveBack_8u_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_8u_C3R, nppiWarpPerspectiveBack_8u_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpPerspective_8u_C4R, nppiWarpPerspectiveBack_8u_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_16u_t npp_warpPerspective_16u[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_16u_C1R, nppiWarpPerspectiveBack_16u_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_16u_C3R, nppiWarpPerspectiveBack_16u_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpPerspective_16u_C4R, nppiWarpPerspectiveBack_16u_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_32s_t npp_warpPerspective_32s[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_32s_C1R, nppiWarpPerspectiveBack_32s_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_32s_C3R, nppiWarpPerspectiveBack_32s_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpPerspective_32s_C4R, nppiWarpPerspectiveBack_32s_C4R}
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
static npp_warp_32f_t npp_warpPerspective_32f[][2] =
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
2010-11-08 17:55:10 +08:00
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_32f_C1R, nppiWarpPerspectiveBack_32f_C1R},
|
|
|
|
{0, 0},
|
|
|
|
{nppiWarpPerspective_32f_C3R, nppiWarpPerspectiveBack_32f_C3R},
|
2010-09-22 18:58:01 +08:00
|
|
|
{nppiWarpPerspective_32f_C4R, nppiWarpPerspectiveBack_32f_C4R}
|
|
|
|
};
|
|
|
|
|
|
|
|
CV_Assert(M.rows == 3 && M.cols == 3);
|
|
|
|
|
|
|
|
double coeffs[3][3];
|
|
|
|
Mat coeffsMat(3, 3, CV_64F, (void*)coeffs);
|
|
|
|
M.convertTo(coeffsMat, coeffsMat.type());
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
nppWarpCaller(src, dst, coeffs, dsize, flags, npp_warpPerspective_8u, npp_warpPerspective_16u, npp_warpPerspective_32s, npp_warpPerspective_32f, StreamAccessor::getStream(s));
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
2011-07-01 15:07:54 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// buildWarpPlaneMaps
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void buildWarpPlaneMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
|
|
|
|
const float r[9], const float rinv[9], float f, float s, float dist,
|
|
|
|
float half_w, float half_h, cudaStream_t stream);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::buildWarpPlaneMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
|
|
|
|
double dist, GpuMat& map_x, GpuMat& map_y, Stream& stream)
|
|
|
|
{
|
|
|
|
CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
|
|
|
|
Mat Rinv = R.inv();
|
|
|
|
CV_Assert(Rinv.isContinuous());
|
|
|
|
|
|
|
|
map_x.create(dst_roi.size(), CV_32F);
|
|
|
|
map_y.create(dst_roi.size(), CV_32F);
|
|
|
|
imgproc::buildWarpPlaneMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
|
2011-08-08 19:28:14 +08:00
|
|
|
static_cast<float>(f), static_cast<float>(s), static_cast<float>(dist),
|
|
|
|
0.5f*src_size.width, 0.5f*src_size.height, StreamAccessor::getStream(stream));
|
2011-07-01 15:07:54 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// buildWarpCylyndricalMaps
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void buildWarpCylindricalMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
|
|
|
|
const float r[9], const float rinv[9], float f, float s,
|
|
|
|
float half_w, float half_h, cudaStream_t stream);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::buildWarpCylindricalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
|
|
|
|
GpuMat& map_x, GpuMat& map_y, Stream& stream)
|
|
|
|
{
|
|
|
|
CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
|
|
|
|
Mat Rinv = R.inv();
|
|
|
|
CV_Assert(Rinv.isContinuous());
|
|
|
|
|
|
|
|
map_x.create(dst_roi.size(), CV_32F);
|
|
|
|
map_y.create(dst_roi.size(), CV_32F);
|
|
|
|
imgproc::buildWarpCylindricalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
|
2011-08-08 19:28:14 +08:00
|
|
|
static_cast<float>(f), static_cast<float>(s), 0.5f*src_size.width, 0.5f*src_size.height,
|
|
|
|
StreamAccessor::getStream(stream));
|
2011-07-01 15:07:54 +08:00
|
|
|
}
|
|
|
|
|
2011-06-30 22:39:48 +08:00
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// buildWarpSphericalMaps
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void buildWarpSphericalMaps(int tl_u, int tl_v, DevMem2Df map_x, DevMem2Df map_y,
|
|
|
|
const float r[9], const float rinv[9], float f, float s,
|
|
|
|
float half_w, float half_h, cudaStream_t stream);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::buildWarpSphericalMaps(Size src_size, Rect dst_roi, const Mat& R, double f, double s,
|
|
|
|
GpuMat& map_x, GpuMat& map_y, Stream& stream)
|
|
|
|
{
|
|
|
|
CV_Assert(R.size() == Size(3,3) && R.isContinuous() && R.type() == CV_32F);
|
|
|
|
Mat Rinv = R.inv();
|
|
|
|
CV_Assert(Rinv.isContinuous());
|
|
|
|
|
|
|
|
map_x.create(dst_roi.size(), CV_32F);
|
|
|
|
map_y.create(dst_roi.size(), CV_32F);
|
|
|
|
imgproc::buildWarpSphericalMaps(dst_roi.tl().x, dst_roi.tl().y, map_x, map_y, R.ptr<float>(), Rinv.ptr<float>(),
|
2011-08-08 19:28:14 +08:00
|
|
|
static_cast<float>(f), static_cast<float>(s), 0.5f*src_size.width, 0.5f*src_size.height,
|
|
|
|
StreamAccessor::getStream(stream));
|
2011-06-30 22:39:48 +08:00
|
|
|
}
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// rotate
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::rotate(const GpuMat& src, GpuMat& dst, Size dsize, double angle, double xShift, double yShift, int interpolation, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
|
|
|
static const int npp_inter[] = {NPPI_INTER_NN, NPPI_INTER_LINEAR, NPPI_INTER_CUBIC};
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_8UC4);
|
|
|
|
CV_Assert(interpolation == INTER_NEAREST || interpolation == INTER_LINEAR || interpolation == INTER_CUBIC);
|
|
|
|
|
|
|
|
dst.create(dsize, src.type());
|
|
|
|
|
|
|
|
NppiSize srcsz;
|
|
|
|
srcsz.height = src.rows;
|
|
|
|
srcsz.width = src.cols;
|
|
|
|
NppiRect srcroi;
|
|
|
|
srcroi.x = srcroi.y = 0;
|
|
|
|
srcroi.height = src.rows;
|
|
|
|
srcroi.width = src.cols;
|
|
|
|
NppiRect dstroi;
|
|
|
|
dstroi.x = dstroi.y = 0;
|
|
|
|
dstroi.height = dst.rows;
|
|
|
|
dstroi.width = dst.cols;
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2010-09-22 18:58:01 +08:00
|
|
|
if (src.type() == CV_8UC1)
|
|
|
|
{
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiRotate_8u_C1R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiRotate_8u_C4R(src.ptr<Npp8u>(), srcsz, static_cast<int>(src.step), srcroi,
|
|
|
|
dst.ptr<Npp8u>(), static_cast<int>(dst.step), dstroi, angle, xShift, yShift, npp_inter[interpolation]) );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// integral
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::integral(const GpuMat& src, GpuMat& sum, Stream& s)
|
2011-01-31 18:42:33 +08:00
|
|
|
{
|
|
|
|
GpuMat buffer;
|
2011-05-31 16:31:10 +08:00
|
|
|
integralBuffered(src, sum, buffer, s);
|
2011-01-31 18:42:33 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::integralBuffered(const GpuMat& src, GpuMat& sum, GpuMat& buffer, Stream& s)
|
2010-12-21 22:02:09 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8UC1);
|
|
|
|
|
|
|
|
sum.create(src.rows + 1, src.cols + 1, CV_32S);
|
|
|
|
|
2011-02-04 23:15:25 +08:00
|
|
|
NcvSize32u roiSize;
|
2010-12-21 22:02:09 +08:00
|
|
|
roiSize.width = src.cols;
|
|
|
|
roiSize.height = src.rows;
|
|
|
|
|
2011-02-04 23:15:25 +08:00
|
|
|
cudaDeviceProp prop;
|
|
|
|
cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
|
|
|
|
|
|
|
|
Ncv32u bufSize;
|
|
|
|
nppSafeCall( nppiStIntegralGetSize_8u32u(roiSize, &bufSize, prop) );
|
2011-01-31 18:42:33 +08:00
|
|
|
ensureSizeIsEnough(1, bufSize, CV_8UC1, buffer);
|
2010-12-21 22:02:09 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiStIntegral_8u32u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>()), static_cast<int>(src.step),
|
|
|
|
sum.ptr<Ncv32u>(), static_cast<int>(sum.step), roiSize, buffer.ptr<Ncv8u>(), bufSize, prop) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-12-21 22:02:09 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::integral(const GpuMat& src, GpuMat& sum, GpuMat& sqsum, Stream& s)
|
2010-09-22 18:58:01 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8UC1);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
int width = src.cols + 1, height = src.rows + 1;
|
2010-09-22 18:58:01 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
sum.create(height, width, CV_32S);
|
|
|
|
sqsum.create(height, width, CV_32F);
|
2010-09-22 18:58:01 +08:00
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiSqrIntegral_8u32s32f_C1R(const_cast<Npp8u*>(src.ptr<Npp8u>()), static_cast<int>(src.step),
|
|
|
|
sum.ptr<Npp32s>(), static_cast<int>(sum.step), sqsum.ptr<Npp32f>(), static_cast<int>(sqsum.step), sz, 0, 0.0f, height) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-09-22 18:58:01 +08:00
|
|
|
}
|
|
|
|
|
2010-12-22 16:17:36 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// sqrIntegral
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::sqrIntegral(const GpuMat& src, GpuMat& sqsum, Stream& s)
|
2010-12-22 16:17:36 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8U);
|
|
|
|
|
2011-02-04 23:15:25 +08:00
|
|
|
NcvSize32u roiSize;
|
2010-12-22 16:17:36 +08:00
|
|
|
roiSize.width = src.cols;
|
|
|
|
roiSize.height = src.rows;
|
|
|
|
|
2011-02-04 23:15:25 +08:00
|
|
|
cudaDeviceProp prop;
|
|
|
|
cudaSafeCall( cudaGetDeviceProperties(&prop, cv::gpu::getDevice()) );
|
|
|
|
|
|
|
|
Ncv32u bufSize;
|
|
|
|
nppSafeCall(nppiStSqrIntegralGetSize_8u64u(roiSize, &bufSize, prop));
|
2010-12-22 16:17:36 +08:00
|
|
|
GpuMat buf(1, bufSize, CV_8U);
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStStreamHandler h(stream);
|
|
|
|
|
2010-12-22 16:17:36 +08:00
|
|
|
sqsum.create(src.rows + 1, src.cols + 1, CV_64F);
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall(nppiStSqrIntegral_8u64u_C1R(const_cast<Ncv8u*>(src.ptr<Ncv8u>(0)), static_cast<int>(src.step),
|
|
|
|
sqsum.ptr<Ncv64u>(0), static_cast<int>(sqsum.step), roiSize, buf.ptr<Ncv8u>(0), bufSize, prop));
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-12-22 16:17:36 +08:00
|
|
|
}
|
|
|
|
|
2010-12-08 23:06:10 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// columnSum
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void columnSum_32F(const DevMem2D src, const DevMem2D dst);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::columnSum(const GpuMat& src, GpuMat& dst)
|
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_32F);
|
|
|
|
|
|
|
|
dst.create(src.size(), CV_32F);
|
|
|
|
imgproc::columnSum_32F(src, dst);
|
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::rectStdDev(const GpuMat& src, const GpuMat& sqr, GpuMat& dst, const Rect& rect, Stream& s)
|
2010-10-13 20:52:39 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_32SC1 && sqr.type() == CV_32FC1);
|
|
|
|
|
|
|
|
dst.create(src.size(), CV_32FC1);
|
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
|
|
|
NppiRect nppRect;
|
|
|
|
nppRect.height = rect.height;
|
|
|
|
nppRect.width = rect.width;
|
|
|
|
nppRect.x = rect.x;
|
|
|
|
nppRect.y = rect.y;
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( nppiRectStdDev_32s32f_C1R(src.ptr<Npp32s>(), static_cast<int>(src.step), sqr.ptr<Npp32f>(), static_cast<int>(sqr.step),
|
|
|
|
dst.ptr<Npp32f>(), static_cast<int>(dst.step), sz, nppRect) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-10-13 20:52:39 +08:00
|
|
|
}
|
|
|
|
|
2010-10-04 19:42:40 +08:00
|
|
|
|
2010-10-12 14:37:44 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// Histogram
|
|
|
|
|
|
|
|
namespace
|
|
|
|
{
|
|
|
|
template<int n> struct NPPTypeTraits;
|
|
|
|
template<> struct NPPTypeTraits<CV_8U> { typedef Npp8u npp_type; };
|
|
|
|
template<> struct NPPTypeTraits<CV_16U> { typedef Npp16u npp_type; };
|
|
|
|
template<> struct NPPTypeTraits<CV_16S> { typedef Npp16s npp_type; };
|
2010-10-13 20:52:39 +08:00
|
|
|
template<> struct NPPTypeTraits<CV_32F> { typedef Npp32f npp_type; };
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-10-12 14:37:44 +08:00
|
|
|
typedef NppStatus (*get_buf_size_c1_t)(NppiSize oSizeROI, int nLevels, int* hpBufferSize);
|
|
|
|
typedef NppStatus (*get_buf_size_c4_t)(NppiSize oSizeROI, int nLevels[], int* hpBufferSize);
|
|
|
|
|
|
|
|
template<int SDEPTH> struct NppHistogramEvenFuncC1
|
|
|
|
{
|
|
|
|
typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s * pHist,
|
2010-10-12 14:37:44 +08:00
|
|
|
int nLevels, Npp32s nLowerLevel, Npp32s nUpperLevel, Npp8u * pBuffer);
|
|
|
|
};
|
|
|
|
template<int SDEPTH> struct NppHistogramEvenFuncC4
|
|
|
|
{
|
|
|
|
typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI,
|
2010-10-12 14:37:44 +08:00
|
|
|
Npp32s * pHist[4], int nLevels[4], Npp32s nLowerLevel[4], Npp32s nUpperLevel[4], Npp8u * pBuffer);
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
|
|
|
|
template<int SDEPTH, typename NppHistogramEvenFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
|
2010-10-12 14:37:44 +08:00
|
|
|
struct NppHistogramEvenC1
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-12 14:37:44 +08:00
|
|
|
typedef typename NppHistogramEvenFuncC1<SDEPTH>::src_t src_t;
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
static void hist(const GpuMat& src, GpuMat& hist, GpuMat& buffer, int histSize, int lowerLevel, int upperLevel, cudaStream_t stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
int levels = histSize + 1;
|
|
|
|
hist.create(1, histSize, CV_32S);
|
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
|
|
|
int buf_size;
|
|
|
|
get_buf_size(sz, levels, &buf_size);
|
2011-08-08 21:05:50 +08:00
|
|
|
|
|
|
|
ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
|
2011-05-31 16:31:10 +08:00
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels,
|
2010-10-12 14:37:44 +08:00
|
|
|
lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
2010-11-08 17:55:10 +08:00
|
|
|
};
|
|
|
|
template<int SDEPTH, typename NppHistogramEvenFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
|
2010-10-12 14:37:44 +08:00
|
|
|
struct NppHistogramEvenC4
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-12 14:37:44 +08:00
|
|
|
typedef typename NppHistogramEvenFuncC4<SDEPTH>::src_t src_t;
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
static void hist(const GpuMat& src, GpuMat hist[4], GpuMat& buffer, int histSize[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
int levels[] = {histSize[0] + 1, histSize[1] + 1, histSize[2] + 1, histSize[3] + 1};
|
|
|
|
hist[0].create(1, histSize[0], CV_32S);
|
|
|
|
hist[1].create(1, histSize[1], CV_32S);
|
|
|
|
hist[2].create(1, histSize[2], CV_32S);
|
|
|
|
hist[3].create(1, histSize[3], CV_32S);
|
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
|
|
|
Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
|
|
|
|
|
|
|
|
int buf_size;
|
|
|
|
get_buf_size(sz, levels, &buf_size);
|
2011-08-08 21:05:50 +08:00
|
|
|
|
|
|
|
ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
|
2011-05-31 16:31:10 +08:00
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, levels, lowerLevel, upperLevel, buffer.ptr<Npp8u>()) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
template<int SDEPTH> struct NppHistogramRangeFuncC1
|
|
|
|
{
|
|
|
|
typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
|
2010-10-13 20:52:39 +08:00
|
|
|
typedef Npp32s level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=CV_32SC1};
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
|
2010-10-12 14:37:44 +08:00
|
|
|
const Npp32s* pLevels, int nLevels, Npp8u* pBuffer);
|
|
|
|
};
|
2010-10-13 20:52:39 +08:00
|
|
|
template<> struct NppHistogramRangeFuncC1<CV_32F>
|
|
|
|
{
|
|
|
|
typedef Npp32f src_t;
|
|
|
|
typedef Npp32f level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=CV_32FC1};
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist,
|
2010-10-13 20:52:39 +08:00
|
|
|
const Npp32f* pLevels, int nLevels, Npp8u* pBuffer);
|
|
|
|
};
|
2010-10-12 14:37:44 +08:00
|
|
|
template<int SDEPTH> struct NppHistogramRangeFuncC4
|
|
|
|
{
|
|
|
|
typedef typename NPPTypeTraits<SDEPTH>::npp_type src_t;
|
2010-10-13 20:52:39 +08:00
|
|
|
typedef Npp32s level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=CV_32SC1};
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const src_t* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
|
2010-10-12 14:37:44 +08:00
|
|
|
const Npp32s* pLevels[4], int nLevels[4], Npp8u* pBuffer);
|
|
|
|
};
|
2010-10-13 20:52:39 +08:00
|
|
|
template<> struct NppHistogramRangeFuncC4<CV_32F>
|
|
|
|
{
|
|
|
|
typedef Npp32f src_t;
|
|
|
|
typedef Npp32f level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=CV_32FC1};
|
|
|
|
|
2010-11-08 17:55:10 +08:00
|
|
|
typedef NppStatus (*func_ptr)(const Npp32f* pSrc, int nSrcStep, NppiSize oSizeROI, Npp32s* pHist[4],
|
2010-10-13 20:52:39 +08:00
|
|
|
const Npp32f* pLevels[4], int nLevels[4], Npp8u* pBuffer);
|
|
|
|
};
|
2010-11-08 17:55:10 +08:00
|
|
|
|
|
|
|
template<int SDEPTH, typename NppHistogramRangeFuncC1<SDEPTH>::func_ptr func, get_buf_size_c1_t get_buf_size>
|
2010-10-12 14:37:44 +08:00
|
|
|
struct NppHistogramRangeC1
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-12 14:37:44 +08:00
|
|
|
typedef typename NppHistogramRangeFuncC1<SDEPTH>::src_t src_t;
|
2010-10-13 20:52:39 +08:00
|
|
|
typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
static void hist(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buffer, cudaStream_t stream)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-13 20:52:39 +08:00
|
|
|
CV_Assert(levels.type() == LEVEL_TYPE_CODE && levels.rows == 1);
|
2010-10-12 14:37:44 +08:00
|
|
|
|
|
|
|
hist.create(1, levels.cols - 1, CV_32S);
|
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
|
|
|
int buf_size;
|
|
|
|
get_buf_size(sz, levels.cols, &buf_size);
|
2011-08-08 21:05:50 +08:00
|
|
|
|
|
|
|
ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
|
2011-05-31 16:31:10 +08:00
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, hist.ptr<Npp32s>(), levels.ptr<level_t>(), levels.cols, buffer.ptr<Npp8u>()) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
2010-11-08 17:55:10 +08:00
|
|
|
};
|
|
|
|
template<int SDEPTH, typename NppHistogramRangeFuncC4<SDEPTH>::func_ptr func, get_buf_size_c4_t get_buf_size>
|
2010-10-12 14:37:44 +08:00
|
|
|
struct NppHistogramRangeC4
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-10-12 14:37:44 +08:00
|
|
|
typedef typename NppHistogramRangeFuncC4<SDEPTH>::src_t src_t;
|
2010-10-13 20:52:39 +08:00
|
|
|
typedef typename NppHistogramRangeFuncC1<SDEPTH>::level_t level_t;
|
|
|
|
enum {LEVEL_TYPE_CODE=NppHistogramRangeFuncC1<SDEPTH>::LEVEL_TYPE_CODE};
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
static void hist(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buffer, cudaStream_t stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
2010-10-13 20:52:39 +08:00
|
|
|
CV_Assert(levels[0].type() == LEVEL_TYPE_CODE && levels[0].rows == 1);
|
|
|
|
CV_Assert(levels[1].type() == LEVEL_TYPE_CODE && levels[1].rows == 1);
|
|
|
|
CV_Assert(levels[2].type() == LEVEL_TYPE_CODE && levels[2].rows == 1);
|
|
|
|
CV_Assert(levels[3].type() == LEVEL_TYPE_CODE && levels[3].rows == 1);
|
2010-10-12 14:37:44 +08:00
|
|
|
|
|
|
|
hist[0].create(1, levels[0].cols - 1, CV_32S);
|
|
|
|
hist[1].create(1, levels[1].cols - 1, CV_32S);
|
|
|
|
hist[2].create(1, levels[2].cols - 1, CV_32S);
|
|
|
|
hist[3].create(1, levels[3].cols - 1, CV_32S);
|
|
|
|
|
|
|
|
Npp32s* pHist[] = {hist[0].ptr<Npp32s>(), hist[1].ptr<Npp32s>(), hist[2].ptr<Npp32s>(), hist[3].ptr<Npp32s>()};
|
|
|
|
int nLevels[] = {levels[0].cols, levels[1].cols, levels[2].cols, levels[3].cols};
|
2010-10-13 20:52:39 +08:00
|
|
|
const level_t* pLevels[] = {levels[0].ptr<level_t>(), levels[1].ptr<level_t>(), levels[2].ptr<level_t>(), levels[3].ptr<level_t>()};
|
2010-10-12 14:37:44 +08:00
|
|
|
|
|
|
|
NppiSize sz;
|
|
|
|
sz.width = src.cols;
|
|
|
|
sz.height = src.rows;
|
|
|
|
|
|
|
|
int buf_size;
|
|
|
|
get_buf_size(sz, nLevels, &buf_size);
|
2011-08-08 21:05:50 +08:00
|
|
|
|
|
|
|
ensureSizeIsEnough(1, buf_size, CV_8U, buffer);
|
2011-05-31 16:31:10 +08:00
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
nppSafeCall( func(src.ptr<src_t>(), static_cast<int>(src.step), sz, pHist, pLevels, nLevels, buffer.ptr<Npp8u>()) );
|
2011-01-24 18:32:57 +08:00
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
2010-11-08 17:55:10 +08:00
|
|
|
};
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::evenLevels(GpuMat& levels, int nLevels, int lowerLevel, int upperLevel)
|
|
|
|
{
|
|
|
|
Mat host_levels(1, nLevels, CV_32SC1);
|
|
|
|
nppSafeCall( nppiEvenLevelsHost_32s(host_levels.ptr<Npp32s>(), nLevels, lowerLevel, upperLevel) );
|
|
|
|
levels.upload(host_levels);
|
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, int histSize, int lowerLevel, int upperLevel, Stream& stream)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::histEven(const GpuMat& src, GpuMat& hist, GpuMat& buf, int histSize, int lowerLevel, int upperLevel, Stream& stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 );
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, GpuMat& buf, int levels, int lowerLevel, int upperLevel, cudaStream_t stream);
|
2010-11-08 17:55:10 +08:00
|
|
|
static const hist_t hist_callers[] =
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
NppHistogramEvenC1<CV_8U , nppiHistogramEven_8u_C1R , nppiHistogramEvenGetBufferSize_8u_C1R >::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramEvenC1<CV_16U, nppiHistogramEven_16u_C1R, nppiHistogramEvenGetBufferSize_16u_C1R>::hist,
|
|
|
|
NppHistogramEvenC1<CV_16S, nppiHistogramEven_16s_C1R, nppiHistogramEvenGetBufferSize_16s_C1R>::hist
|
|
|
|
};
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
histEven(src, hist, buf, histSize, lowerLevel, upperLevel, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::histEven(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int histSize[4], int lowerLevel[4], int upperLevel[4], Stream& stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 );
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], GpuMat& buf, int levels[4], int lowerLevel[4], int upperLevel[4], cudaStream_t stream);
|
2010-11-08 17:55:10 +08:00
|
|
|
static const hist_t hist_callers[] =
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
NppHistogramEvenC4<CV_8U , nppiHistogramEven_8u_C4R , nppiHistogramEvenGetBufferSize_8u_C4R >::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramEvenC4<CV_16U, nppiHistogramEven_16u_C4R, nppiHistogramEvenGetBufferSize_16u_C4R>::hist,
|
|
|
|
NppHistogramEvenC4<CV_16S, nppiHistogramEven_16s_C4R, nppiHistogramEvenGetBufferSize_16s_C4R>::hist
|
|
|
|
};
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
hist_callers[src.depth()](src, hist, buf, histSize, lowerLevel, upperLevel, StreamAccessor::getStream(stream));
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, Stream& stream)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
histRange(src, hist, levels, buf, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void cv::gpu::histRange(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, Stream& stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
2010-10-13 20:52:39 +08:00
|
|
|
CV_Assert(src.type() == CV_8UC1 || src.type() == CV_16UC1 || src.type() == CV_16SC1 || src.type() == CV_32FC1);
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
typedef void (*hist_t)(const GpuMat& src, GpuMat& hist, const GpuMat& levels, GpuMat& buf, cudaStream_t stream);
|
2010-11-08 17:55:10 +08:00
|
|
|
static const hist_t hist_callers[] =
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
NppHistogramRangeC1<CV_8U , nppiHistogramRange_8u_C1R , nppiHistogramRangeGetBufferSize_8u_C1R >::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramRangeC1<CV_16U, nppiHistogramRange_16u_C1R, nppiHistogramRangeGetBufferSize_16u_C1R>::hist,
|
2010-10-13 20:52:39 +08:00
|
|
|
NppHistogramRangeC1<CV_16S, nppiHistogramRange_16s_C1R, nppiHistogramRangeGetBufferSize_16s_C1R>::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramRangeC1<CV_32F, nppiHistogramRange_32f_C1R, nppiHistogramRangeGetBufferSize_32f_C1R>::hist
|
2010-10-12 14:37:44 +08:00
|
|
|
};
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
|
2011-05-31 16:31:10 +08:00
|
|
|
void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], Stream& stream)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
histRange(src, hist, levels, buf, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::histRange(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, Stream& stream)
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
2010-10-13 20:52:39 +08:00
|
|
|
CV_Assert(src.type() == CV_8UC4 || src.type() == CV_16UC4 || src.type() == CV_16SC4 || src.type() == CV_32FC4);
|
2010-10-12 14:37:44 +08:00
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
typedef void (*hist_t)(const GpuMat& src, GpuMat hist[4], const GpuMat levels[4], GpuMat& buf, cudaStream_t stream);
|
2010-11-08 17:55:10 +08:00
|
|
|
static const hist_t hist_callers[] =
|
2010-10-12 14:37:44 +08:00
|
|
|
{
|
|
|
|
NppHistogramRangeC4<CV_8U , nppiHistogramRange_8u_C4R , nppiHistogramRangeGetBufferSize_8u_C4R >::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramRangeC4<CV_16U, nppiHistogramRange_16u_C4R, nppiHistogramRangeGetBufferSize_16u_C4R>::hist,
|
2010-10-13 20:52:39 +08:00
|
|
|
NppHistogramRangeC4<CV_16S, nppiHistogramRange_16s_C4R, nppiHistogramRangeGetBufferSize_16s_C4R>::hist,
|
|
|
|
0,
|
|
|
|
NppHistogramRangeC4<CV_32F, nppiHistogramRange_32f_C4R, nppiHistogramRangeGetBufferSize_32f_C4R>::hist
|
2010-10-12 14:37:44 +08:00
|
|
|
};
|
|
|
|
|
2011-08-08 21:05:50 +08:00
|
|
|
hist_callers[src.depth()](src, hist, levels, buf, StreamAccessor::getStream(stream));
|
2010-10-12 14:37:44 +08:00
|
|
|
}
|
|
|
|
|
2011-08-01 21:15:37 +08:00
|
|
|
namespace cv { namespace gpu { namespace histograms
|
|
|
|
{
|
|
|
|
void histogram256_gpu(DevMem2D src, int* hist, unsigned int* buf, cudaStream_t stream);
|
|
|
|
|
|
|
|
const int PARTIAL_HISTOGRAM256_COUNT = 240;
|
|
|
|
const int HISTOGRAM256_BIN_COUNT = 256;
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::calcHist(const GpuMat& src, GpuMat& hist, Stream& stream)
|
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
calcHist(src, hist, buf, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::calcHist(const GpuMat& src, GpuMat& hist, GpuMat& buf, Stream& stream)
|
|
|
|
{
|
|
|
|
using namespace cv::gpu::histograms;
|
|
|
|
|
|
|
|
CV_Assert(src.type() == CV_8UC1);
|
|
|
|
|
|
|
|
hist.create(1, 256, CV_32SC1);
|
|
|
|
|
|
|
|
ensureSizeIsEnough(1, PARTIAL_HISTOGRAM256_COUNT * HISTOGRAM256_BIN_COUNT, CV_32SC1, buf);
|
|
|
|
|
|
|
|
histogram256_gpu(src, hist.ptr<int>(), buf.ptr<unsigned int>(), StreamAccessor::getStream(stream));
|
|
|
|
}
|
|
|
|
|
2011-08-02 16:33:27 +08:00
|
|
|
void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, Stream& stream)
|
|
|
|
{
|
|
|
|
GpuMat hist;
|
|
|
|
GpuMat buf;
|
|
|
|
equalizeHist(src, dst, hist, buf, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, Stream& stream)
|
|
|
|
{
|
|
|
|
GpuMat buf;
|
|
|
|
equalizeHist(src, dst, hist, buf, stream);
|
|
|
|
}
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace histograms
|
|
|
|
{
|
|
|
|
void equalizeHist_gpu(DevMem2D src, DevMem2D dst, const int* lut, cudaStream_t stream);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
void cv::gpu::equalizeHist(const GpuMat& src, GpuMat& dst, GpuMat& hist, GpuMat& buf, Stream& s)
|
|
|
|
{
|
|
|
|
using namespace cv::gpu::histograms;
|
|
|
|
|
|
|
|
CV_Assert(src.type() == CV_8UC1);
|
|
|
|
|
|
|
|
dst.create(src.size(), src.type());
|
|
|
|
|
|
|
|
int intBufSize;
|
|
|
|
nppSafeCall( nppsIntegralGetBufferSize_32s(256, &intBufSize) );
|
|
|
|
|
2011-08-08 19:28:14 +08:00
|
|
|
int bufSize = static_cast<int>(std::max(256 * 240 * sizeof(int), intBufSize + 256 * sizeof(int)));
|
2011-08-02 16:33:27 +08:00
|
|
|
|
|
|
|
ensureSizeIsEnough(1, bufSize, CV_8UC1, buf);
|
|
|
|
|
|
|
|
GpuMat histBuf(1, 256 * 240, CV_32SC1, buf.ptr());
|
|
|
|
GpuMat intBuf(1, intBufSize, CV_8UC1, buf.ptr());
|
|
|
|
GpuMat lut(1, 256, CV_32S, buf.ptr() + intBufSize);
|
|
|
|
|
|
|
|
calcHist(src, hist, histBuf, s);
|
|
|
|
|
|
|
|
cudaStream_t stream = StreamAccessor::getStream(s);
|
|
|
|
|
|
|
|
NppStreamHandler h(stream);
|
|
|
|
|
|
|
|
nppSafeCall( nppsIntegral_32s(hist.ptr<Npp32s>(), lut.ptr<Npp32s>(), 256, intBuf.ptr<Npp8u>()) );
|
|
|
|
|
|
|
|
if (stream == 0)
|
|
|
|
cudaSafeCall( cudaDeviceSynchronize() );
|
|
|
|
|
|
|
|
equalizeHist_gpu(src, dst, lut.ptr<int>(), stream);
|
|
|
|
}
|
|
|
|
|
2010-12-07 00:37:32 +08:00
|
|
|
////////////////////////////////////////////////////////////////////////
|
|
|
|
// cornerHarris & minEgenVal
|
|
|
|
|
2010-11-30 16:04:37 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc {
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
void extractCovData_caller(const DevMem2Df Dx, const DevMem2Df Dy, PtrStepf dst);
|
|
|
|
void cornerHarris_caller(const int block_size, const float k, const DevMem2D Dx, const DevMem2D Dy, DevMem2D dst, int border_type);
|
|
|
|
void cornerMinEigenVal_caller(const int block_size, const DevMem2D Dx, const DevMem2D Dy, DevMem2D dst, int border_type);
|
|
|
|
|
|
|
|
}}}
|
|
|
|
|
2010-11-30 16:44:04 +08:00
|
|
|
namespace
|
2010-11-30 16:04:37 +08:00
|
|
|
{
|
2010-12-02 17:07:13 +08:00
|
|
|
template <typename T>
|
2011-01-31 21:31:59 +08:00
|
|
|
void extractCovData(const GpuMat& src, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
|
2010-12-02 17:07:13 +08:00
|
|
|
{
|
2010-11-30 16:44:04 +08:00
|
|
|
double scale = (double)(1 << ((ksize > 0 ? ksize : 3) - 1)) * blockSize;
|
2010-12-02 17:07:13 +08:00
|
|
|
if (ksize < 0)
|
|
|
|
scale *= 2.;
|
|
|
|
if (src.depth() == CV_8U)
|
|
|
|
scale *= 255.;
|
2010-11-30 16:44:04 +08:00
|
|
|
scale = 1./scale;
|
2010-11-30 16:04:37 +08:00
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
Dx.create(src.size(), CV_32F);
|
|
|
|
Dy.create(src.size(), CV_32F);
|
|
|
|
|
2011-01-31 21:31:59 +08:00
|
|
|
if (ksize > 0)
|
|
|
|
{
|
|
|
|
Sobel(src, Dx, CV_32F, 1, 0, ksize, scale, borderType);
|
|
|
|
Sobel(src, Dy, CV_32F, 0, 1, ksize, scale, borderType);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
Scharr(src, Dx, CV_32F, 1, 0, scale, borderType);
|
|
|
|
Scharr(src, Dy, CV_32F, 0, 1, scale, borderType);
|
|
|
|
}
|
2010-12-02 17:07:13 +08:00
|
|
|
}
|
|
|
|
|
2011-01-31 21:31:59 +08:00
|
|
|
void extractCovData(const GpuMat& src, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
|
2010-12-02 17:07:13 +08:00
|
|
|
{
|
|
|
|
switch (src.type())
|
2010-11-30 16:44:04 +08:00
|
|
|
{
|
2010-12-02 17:07:13 +08:00
|
|
|
case CV_8U:
|
2011-01-31 21:31:59 +08:00
|
|
|
extractCovData<unsigned char>(src, Dx, Dy, blockSize, ksize, borderType);
|
2010-12-02 17:07:13 +08:00
|
|
|
break;
|
|
|
|
case CV_32F:
|
2011-01-31 21:31:59 +08:00
|
|
|
extractCovData<float>(src, Dx, Dy, blockSize, ksize, borderType);
|
2010-12-02 17:07:13 +08:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
CV_Error(CV_StsBadArg, "extractCovData: unsupported type of the source matrix");
|
2010-11-30 16:44:04 +08:00
|
|
|
}
|
2010-11-30 16:04:37 +08:00
|
|
|
}
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
} // Anonymous namespace
|
|
|
|
|
2010-12-07 00:37:32 +08:00
|
|
|
|
|
|
|
bool cv::gpu::tryConvertToGpuBorderType(int cpuBorderType, int& gpuBorderType)
|
|
|
|
{
|
2011-08-31 21:21:14 +08:00
|
|
|
switch (cpuBorderType)
|
2010-12-07 00:37:32 +08:00
|
|
|
{
|
2011-08-31 21:21:14 +08:00
|
|
|
case cv::BORDER_REFLECT101:
|
2010-12-07 00:37:32 +08:00
|
|
|
gpuBorderType = cv::gpu::BORDER_REFLECT101_GPU;
|
|
|
|
return true;
|
2011-08-31 21:21:14 +08:00
|
|
|
case cv::BORDER_REPLICATE:
|
2010-12-07 00:37:32 +08:00
|
|
|
gpuBorderType = cv::gpu::BORDER_REPLICATE_GPU;
|
|
|
|
return true;
|
2011-08-31 21:21:14 +08:00
|
|
|
case cv::BORDER_CONSTANT:
|
2010-12-13 16:43:04 +08:00
|
|
|
gpuBorderType = cv::gpu::BORDER_CONSTANT_GPU;
|
|
|
|
return true;
|
2011-08-31 21:21:14 +08:00
|
|
|
case cv::BORDER_REFLECT:
|
|
|
|
gpuBorderType = cv::gpu::BORDER_REFLECT_GPU;
|
|
|
|
return true;
|
|
|
|
case cv::BORDER_WRAP:
|
|
|
|
gpuBorderType = cv::gpu::BORDER_WRAP_GPU;
|
|
|
|
return true;
|
|
|
|
default:
|
|
|
|
return false;
|
|
|
|
};
|
2010-12-07 00:37:32 +08:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, double k, int borderType)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat Dx, Dy;
|
|
|
|
cornerHarris(src, dst, Dx, Dy, blockSize, ksize, k, borderType);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::cornerHarris(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, double k, int borderType)
|
2010-11-30 16:44:04 +08:00
|
|
|
{
|
2010-12-06 17:44:51 +08:00
|
|
|
CV_Assert(borderType == cv::BORDER_REFLECT101 ||
|
|
|
|
borderType == cv::BORDER_REPLICATE);
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
int gpuBorderType;
|
2010-12-06 17:44:51 +08:00
|
|
|
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
|
2010-12-02 17:07:13 +08:00
|
|
|
|
2011-01-31 21:31:59 +08:00
|
|
|
extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
|
2010-11-30 16:04:37 +08:00
|
|
|
dst.create(src.size(), CV_32F);
|
2010-12-02 17:07:13 +08:00
|
|
|
imgproc::cornerHarris_caller(blockSize, (float)k, Dx, Dy, dst, gpuBorderType);
|
2010-11-30 16:04:37 +08:00
|
|
|
}
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, int blockSize, int ksize, int borderType)
|
2011-08-08 21:05:50 +08:00
|
|
|
{
|
|
|
|
GpuMat Dx, Dy;
|
|
|
|
cornerMinEigenVal(src, dst, Dx, Dy, blockSize, ksize, borderType);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::cornerMinEigenVal(const GpuMat& src, GpuMat& dst, GpuMat& Dx, GpuMat& Dy, int blockSize, int ksize, int borderType)
|
2010-12-02 17:07:13 +08:00
|
|
|
{
|
2010-12-06 17:44:51 +08:00
|
|
|
CV_Assert(borderType == cv::BORDER_REFLECT101 ||
|
|
|
|
borderType == cv::BORDER_REPLICATE);
|
|
|
|
|
2010-12-02 17:07:13 +08:00
|
|
|
int gpuBorderType;
|
2010-12-06 17:44:51 +08:00
|
|
|
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
|
2010-12-02 17:07:13 +08:00
|
|
|
|
2011-01-31 21:31:59 +08:00
|
|
|
extractCovData(src, Dx, Dy, blockSize, ksize, borderType);
|
2010-11-30 16:44:04 +08:00
|
|
|
dst.create(src.size(), CV_32F);
|
2010-12-02 17:07:13 +08:00
|
|
|
imgproc::cornerMinEigenVal_caller(blockSize, Dx, Dy, dst, gpuBorderType);
|
2010-11-30 16:44:04 +08:00
|
|
|
}
|
|
|
|
|
2010-12-22 21:46:06 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// mulSpectrums
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void mulSpectrums(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
|
|
|
|
DevMem2D_<cufftComplex> c);
|
|
|
|
|
|
|
|
void mulSpectrums_CONJ(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
|
|
|
|
DevMem2D_<cufftComplex> c);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
|
|
|
|
void cv::gpu::mulSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c,
|
|
|
|
int flags, bool conjB)
|
|
|
|
{
|
|
|
|
typedef void (*Caller)(const PtrStep_<cufftComplex>, const PtrStep_<cufftComplex>,
|
|
|
|
DevMem2D_<cufftComplex>);
|
|
|
|
static Caller callers[] = { imgproc::mulSpectrums,
|
|
|
|
imgproc::mulSpectrums_CONJ };
|
|
|
|
|
|
|
|
CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
|
|
|
|
CV_Assert(a.size() == b.size());
|
|
|
|
|
|
|
|
c.create(a.size(), CV_32FC2);
|
|
|
|
|
|
|
|
Caller caller = callers[(int)conjB];
|
|
|
|
caller(a, b, c);
|
|
|
|
}
|
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// mulAndScaleSpectrums
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
void mulAndScaleSpectrums(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
|
|
|
|
float scale, DevMem2D_<cufftComplex> c);
|
|
|
|
|
|
|
|
void mulAndScaleSpectrums_CONJ(const PtrStep_<cufftComplex> a, const PtrStep_<cufftComplex> b,
|
|
|
|
float scale, DevMem2D_<cufftComplex> c);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
|
|
|
|
void cv::gpu::mulAndScaleSpectrums(const GpuMat& a, const GpuMat& b, GpuMat& c,
|
|
|
|
int flags, float scale, bool conjB)
|
|
|
|
{
|
|
|
|
typedef void (*Caller)(const PtrStep_<cufftComplex>, const PtrStep_<cufftComplex>,
|
|
|
|
float scale, DevMem2D_<cufftComplex>);
|
|
|
|
static Caller callers[] = { imgproc::mulAndScaleSpectrums,
|
|
|
|
imgproc::mulAndScaleSpectrums_CONJ };
|
|
|
|
|
|
|
|
CV_Assert(a.type() == b.type() && a.type() == CV_32FC2);
|
|
|
|
CV_Assert(a.size() == b.size());
|
|
|
|
|
|
|
|
c.create(a.size(), CV_32FC2);
|
|
|
|
|
|
|
|
Caller caller = callers[(int)conjB];
|
|
|
|
caller(a, b, scale, c);
|
|
|
|
}
|
|
|
|
|
2010-12-23 17:24:33 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// dft
|
|
|
|
|
2010-12-27 15:35:41 +08:00
|
|
|
void cv::gpu::dft(const GpuMat& src, GpuMat& dst, Size dft_size, int flags)
|
2010-12-23 17:24:33 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.type() == CV_32F || src.type() == CV_32FC2);
|
|
|
|
|
|
|
|
// We don't support unpacked output (in the case of real input)
|
|
|
|
CV_Assert(!(flags & DFT_COMPLEX_OUTPUT));
|
|
|
|
|
2010-12-27 15:35:41 +08:00
|
|
|
bool is_1d_input = (dft_size.height == 1) || (dft_size.width == 1);
|
2010-12-23 17:24:33 +08:00
|
|
|
int is_row_dft = flags & DFT_ROWS;
|
|
|
|
int is_scaled_dft = flags & DFT_SCALE;
|
|
|
|
int is_inverse = flags & DFT_INVERSE;
|
|
|
|
bool is_complex_input = src.channels() == 2;
|
|
|
|
bool is_complex_output = !(flags & DFT_REAL_OUTPUT);
|
|
|
|
|
|
|
|
// We don't support real-to-real transform
|
|
|
|
CV_Assert(is_complex_input || is_complex_output);
|
|
|
|
|
2010-12-24 17:26:19 +08:00
|
|
|
GpuMat src_data;
|
2010-12-23 17:24:33 +08:00
|
|
|
|
|
|
|
// Make sure here we work with the continuous input,
|
|
|
|
// as CUFFT can't handle gaps
|
2010-12-24 17:26:19 +08:00
|
|
|
src_data = src;
|
|
|
|
createContinuous(src.rows, src.cols, src.type(), src_data);
|
|
|
|
if (src_data.data != src.data)
|
|
|
|
src.copyTo(src_data);
|
2010-12-23 17:24:33 +08:00
|
|
|
|
2010-12-27 17:10:22 +08:00
|
|
|
Size dft_size_opt = dft_size;
|
2010-12-24 17:26:19 +08:00
|
|
|
if (is_1d_input && !is_row_dft)
|
2010-12-27 15:35:41 +08:00
|
|
|
{
|
|
|
|
// If the source matrix is single column handle it as single row
|
2010-12-27 17:10:22 +08:00
|
|
|
dft_size_opt.width = std::max(dft_size.width, dft_size.height);
|
|
|
|
dft_size_opt.height = std::min(dft_size.width, dft_size.height);
|
2010-12-27 15:35:41 +08:00
|
|
|
}
|
2010-12-23 17:24:33 +08:00
|
|
|
|
|
|
|
cufftType dft_type = CUFFT_R2C;
|
|
|
|
if (is_complex_input)
|
|
|
|
dft_type = is_complex_output ? CUFFT_C2C : CUFFT_C2R;
|
|
|
|
|
2010-12-27 17:10:22 +08:00
|
|
|
CV_Assert(dft_size_opt.width > 1);
|
2010-12-23 17:24:33 +08:00
|
|
|
|
|
|
|
cufftHandle plan;
|
|
|
|
if (is_1d_input || is_row_dft)
|
2010-12-27 17:10:22 +08:00
|
|
|
cufftPlan1d(&plan, dft_size_opt.width, dft_type, dft_size_opt.height);
|
2010-12-23 17:24:33 +08:00
|
|
|
else
|
2010-12-27 17:10:22 +08:00
|
|
|
cufftPlan2d(&plan, dft_size_opt.height, dft_size_opt.width, dft_type);
|
2010-12-23 17:24:33 +08:00
|
|
|
|
|
|
|
if (is_complex_input)
|
|
|
|
{
|
|
|
|
if (is_complex_output)
|
|
|
|
{
|
2010-12-27 15:35:41 +08:00
|
|
|
createContinuous(dft_size, CV_32FC2, dst);
|
2010-12-23 17:24:33 +08:00
|
|
|
cufftSafeCall(cufftExecC2C(
|
2010-12-24 17:26:19 +08:00
|
|
|
plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftComplex>(),
|
2010-12-23 17:24:33 +08:00
|
|
|
is_inverse ? CUFFT_INVERSE : CUFFT_FORWARD));
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2010-12-27 15:35:41 +08:00
|
|
|
createContinuous(dft_size, CV_32F, dst);
|
2010-12-23 17:24:33 +08:00
|
|
|
cufftSafeCall(cufftExecC2R(
|
2010-12-24 17:26:19 +08:00
|
|
|
plan, src_data.ptr<cufftComplex>(), dst.ptr<cufftReal>()));
|
2010-12-23 17:24:33 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2010-12-27 17:10:22 +08:00
|
|
|
// We could swap dft_size for efficiency. Here we must reflect it
|
|
|
|
if (dft_size == dft_size_opt)
|
2010-12-27 15:35:41 +08:00
|
|
|
createContinuous(Size(dft_size.width / 2 + 1, dft_size.height), CV_32FC2, dst);
|
|
|
|
else
|
|
|
|
createContinuous(Size(dft_size.width, dft_size.height / 2 + 1), CV_32FC2, dst);
|
2010-12-23 17:24:33 +08:00
|
|
|
|
|
|
|
cufftSafeCall(cufftExecR2C(
|
2010-12-24 17:26:19 +08:00
|
|
|
plan, src_data.ptr<cufftReal>(), dst.ptr<cufftComplex>()));
|
2010-12-23 17:24:33 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
cufftSafeCall(cufftDestroy(plan));
|
2010-12-24 14:48:23 +08:00
|
|
|
|
|
|
|
if (is_scaled_dft)
|
2010-12-27 17:10:22 +08:00
|
|
|
multiply(dst, Scalar::all(1. / dft_size.area()), dst);
|
2010-12-23 17:24:33 +08:00
|
|
|
}
|
|
|
|
|
2010-12-22 16:56:16 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
2010-12-27 18:18:42 +08:00
|
|
|
// convolve
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
|
|
|
|
void cv::gpu::ConvolveBuf::create(Size image_size, Size templ_size)
|
2010-12-22 16:56:16 +08:00
|
|
|
{
|
2010-12-27 18:18:42 +08:00
|
|
|
result_size = Size(image_size.width - templ_size.width + 1,
|
|
|
|
image_size.height - templ_size.height + 1);
|
|
|
|
block_size = estimateBlockSize(result_size, templ_size);
|
|
|
|
|
|
|
|
dft_size.width = getOptimalDFTSize(block_size.width + templ_size.width - 1);
|
|
|
|
dft_size.height = getOptimalDFTSize(block_size.width + templ_size.height - 1);
|
|
|
|
createContinuous(dft_size, CV_32F, image_block);
|
|
|
|
createContinuous(dft_size, CV_32F, templ_block);
|
|
|
|
createContinuous(dft_size, CV_32F, result_data);
|
|
|
|
|
|
|
|
spect_len = dft_size.height * (dft_size.width / 2 + 1);
|
|
|
|
createContinuous(1, spect_len, CV_32FC2, image_spect);
|
|
|
|
createContinuous(1, spect_len, CV_32FC2, templ_spect);
|
|
|
|
createContinuous(1, spect_len, CV_32FC2, result_spect);
|
|
|
|
|
|
|
|
block_size.width = std::min(dft_size.width - templ_size.width + 1, result_size.width);
|
|
|
|
block_size.height = std::min(dft_size.height - templ_size.height + 1, result_size.height);
|
|
|
|
}
|
2010-12-22 16:56:16 +08:00
|
|
|
|
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
Size cv::gpu::ConvolveBuf::estimateBlockSize(Size result_size, Size templ_size)
|
|
|
|
{
|
|
|
|
int scale = 40;
|
|
|
|
Size bsize_min(1024, 1024);
|
|
|
|
|
|
|
|
// Check whether we use Fermi generation or newer GPU
|
2011-02-03 22:51:58 +08:00
|
|
|
if (DeviceInfo().majorVersion() >= 2)
|
2010-12-27 18:18:42 +08:00
|
|
|
{
|
|
|
|
bsize_min.width = 2048;
|
|
|
|
bsize_min.height = 2048;
|
2010-12-22 16:56:16 +08:00
|
|
|
}
|
2010-12-27 18:18:42 +08:00
|
|
|
|
|
|
|
Size bsize(std::max(templ_size.width * scale, bsize_min.width),
|
|
|
|
std::max(templ_size.height * scale, bsize_min.height));
|
|
|
|
|
|
|
|
bsize.width = std::min(bsize.width, result_size.width);
|
|
|
|
bsize.height = std::min(bsize.height, result_size.height);
|
|
|
|
return bsize;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
|
|
|
|
bool ccorr)
|
|
|
|
{
|
|
|
|
ConvolveBuf buf;
|
|
|
|
convolve(image, templ, result, ccorr, buf);
|
2010-12-22 16:56:16 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
void cv::gpu::convolve(const GpuMat& image, const GpuMat& templ, GpuMat& result,
|
|
|
|
bool ccorr, ConvolveBuf& buf)
|
2010-12-22 16:56:16 +08:00
|
|
|
{
|
2010-12-22 21:46:06 +08:00
|
|
|
StaticAssert<sizeof(float) == sizeof(cufftReal)>::check();
|
|
|
|
StaticAssert<sizeof(float) * 2 == sizeof(cufftComplex)>::check();
|
2010-12-22 16:56:16 +08:00
|
|
|
|
|
|
|
CV_Assert(image.type() == CV_32F);
|
|
|
|
CV_Assert(templ.type() == CV_32F);
|
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
buf.create(image.size(), templ.size());
|
|
|
|
result.create(buf.result_size, CV_32F);
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
Size& block_size = buf.block_size;
|
|
|
|
Size& dft_size = buf.dft_size;
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
GpuMat& image_block = buf.image_block;
|
|
|
|
GpuMat& templ_block = buf.templ_block;
|
|
|
|
GpuMat& result_data = buf.result_data;
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-27 18:18:42 +08:00
|
|
|
GpuMat& image_spect = buf.image_spect;
|
|
|
|
GpuMat& templ_spect = buf.templ_spect;
|
|
|
|
GpuMat& result_spect = buf.result_spect;
|
2010-12-22 16:56:16 +08:00
|
|
|
|
|
|
|
cufftHandle planR2C, planC2R;
|
|
|
|
cufftSafeCall(cufftPlan2d(&planC2R, dft_size.height, dft_size.width, CUFFT_C2R));
|
|
|
|
cufftSafeCall(cufftPlan2d(&planR2C, dft_size.height, dft_size.width, CUFFT_R2C));
|
|
|
|
|
2010-12-22 21:46:06 +08:00
|
|
|
GpuMat templ_roi(templ.size(), CV_32F, templ.data, templ.step);
|
2010-12-22 16:56:16 +08:00
|
|
|
copyMakeBorder(templ_roi, templ_block, 0, templ_block.rows - templ_roi.rows, 0,
|
|
|
|
templ_block.cols - templ_roi.cols, 0);
|
|
|
|
|
2010-12-24 17:26:19 +08:00
|
|
|
cufftSafeCall(cufftExecR2C(planR2C, templ_block.ptr<cufftReal>(),
|
2010-12-22 21:46:06 +08:00
|
|
|
templ_spect.ptr<cufftComplex>()));
|
2010-12-22 16:56:16 +08:00
|
|
|
|
|
|
|
// Process all blocks of the result matrix
|
|
|
|
for (int y = 0; y < result.rows; y += block_size.height)
|
|
|
|
{
|
|
|
|
for (int x = 0; x < result.cols; x += block_size.width)
|
2010-12-24 20:55:43 +08:00
|
|
|
{
|
|
|
|
Size image_roi_size(std::min(x + dft_size.width, image.cols) - x,
|
|
|
|
std::min(y + dft_size.height, image.rows) - y);
|
2010-12-27 18:18:42 +08:00
|
|
|
GpuMat image_roi(image_roi_size, CV_32F, (void*)(image.ptr<float>(y) + x),
|
|
|
|
image.step);
|
|
|
|
copyMakeBorder(image_roi, image_block, 0, image_block.rows - image_roi.rows,
|
|
|
|
0, image_block.cols - image_roi.cols, 0);
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-24 17:26:19 +08:00
|
|
|
cufftSafeCall(cufftExecR2C(planR2C, image_block.ptr<cufftReal>(),
|
2010-12-22 21:46:06 +08:00
|
|
|
image_spect.ptr<cufftComplex>()));
|
|
|
|
mulAndScaleSpectrums(image_spect, templ_spect, result_spect, 0,
|
2010-12-24 17:26:19 +08:00
|
|
|
1.f / dft_size.area(), ccorr);
|
2010-12-22 21:46:06 +08:00
|
|
|
cufftSafeCall(cufftExecC2R(planC2R, result_spect.ptr<cufftComplex>(),
|
|
|
|
result_data.ptr<cufftReal>()));
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2010-12-24 20:55:43 +08:00
|
|
|
Size result_roi_size(std::min(x + block_size.width, result.cols) - x,
|
|
|
|
std::min(y + block_size.height, result.rows) - y);
|
2010-12-27 18:18:42 +08:00
|
|
|
GpuMat result_roi(result_roi_size, result.type(),
|
|
|
|
(void*)(result.ptr<float>(y) + x), result.step);
|
|
|
|
GpuMat result_block(result_roi_size, result_data.type(),
|
|
|
|
result_data.ptr(), result_data.step);
|
2010-12-22 16:56:16 +08:00
|
|
|
result_block.copyTo(result_roi);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
cufftSafeCall(cufftDestroy(planR2C));
|
|
|
|
cufftSafeCall(cufftDestroy(planC2R));
|
|
|
|
}
|
|
|
|
|
2011-06-30 22:39:48 +08:00
|
|
|
|
2011-04-08 16:04:56 +08:00
|
|
|
////////////////////////////////////////////////////////////////////
|
|
|
|
// downsample
|
2010-12-22 16:56:16 +08:00
|
|
|
|
2011-04-08 16:04:56 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
2011-06-30 22:39:48 +08:00
|
|
|
template <typename T, int cn>
|
2011-08-01 16:15:31 +08:00
|
|
|
void downsampleCaller(const DevMem2D src, DevMem2D dst, cudaStream_t stream);
|
2011-04-08 16:04:56 +08:00
|
|
|
}}}
|
|
|
|
|
2011-06-30 22:39:48 +08:00
|
|
|
|
2011-08-01 16:15:31 +08:00
|
|
|
void cv::gpu::downsample(const GpuMat& src, GpuMat& dst, Stream& stream)
|
2011-04-08 16:04:56 +08:00
|
|
|
{
|
2011-06-30 22:39:48 +08:00
|
|
|
CV_Assert(src.depth() < CV_64F && src.channels() <= 4);
|
|
|
|
|
2011-08-01 16:15:31 +08:00
|
|
|
typedef void (*Caller)(const DevMem2D, DevMem2D, cudaStream_t stream);
|
2011-06-30 22:39:48 +08:00
|
|
|
static const Caller callers[6][4] =
|
|
|
|
{{imgproc::downsampleCaller<uchar,1>, imgproc::downsampleCaller<uchar,2>,
|
|
|
|
imgproc::downsampleCaller<uchar,3>, imgproc::downsampleCaller<uchar,4>},
|
|
|
|
{0,0,0,0}, {0,0,0,0},
|
|
|
|
{imgproc::downsampleCaller<short,1>, imgproc::downsampleCaller<short,2>,
|
|
|
|
imgproc::downsampleCaller<short,3>, imgproc::downsampleCaller<short,4>},
|
|
|
|
{0,0,0,0},
|
|
|
|
{imgproc::downsampleCaller<float,1>, imgproc::downsampleCaller<float,2>,
|
|
|
|
imgproc::downsampleCaller<float,3>, imgproc::downsampleCaller<float,4>}};
|
|
|
|
|
|
|
|
Caller caller = callers[src.depth()][src.channels()-1];
|
|
|
|
if (!caller)
|
|
|
|
CV_Error(CV_StsUnsupportedFormat, "bad number of channels");
|
|
|
|
|
|
|
|
dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
|
2011-08-01 16:15:31 +08:00
|
|
|
caller(src, dst.reshape(1), StreamAccessor::getStream(stream));
|
2011-06-30 22:39:48 +08:00
|
|
|
}
|
2011-04-08 16:04:56 +08:00
|
|
|
|
|
|
|
|
2011-06-30 22:39:48 +08:00
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// upsample
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
|
|
|
{
|
|
|
|
template <typename T, int cn>
|
2011-08-01 16:15:31 +08:00
|
|
|
void upsampleCaller(const DevMem2D src, DevMem2D dst, cudaStream_t stream);
|
2011-06-30 22:39:48 +08:00
|
|
|
}}}
|
|
|
|
|
|
|
|
|
2011-08-01 16:15:31 +08:00
|
|
|
void cv::gpu::upsample(const GpuMat& src, GpuMat& dst, Stream& stream)
|
2011-06-30 22:39:48 +08:00
|
|
|
{
|
|
|
|
CV_Assert(src.depth() < CV_64F && src.channels() <= 4);
|
|
|
|
|
2011-08-01 16:15:31 +08:00
|
|
|
typedef void (*Caller)(const DevMem2D, DevMem2D, cudaStream_t stream);
|
2011-06-30 22:39:48 +08:00
|
|
|
static const Caller callers[6][5] =
|
|
|
|
{{imgproc::upsampleCaller<uchar,1>, imgproc::upsampleCaller<uchar,2>,
|
|
|
|
imgproc::upsampleCaller<uchar,3>, imgproc::upsampleCaller<uchar,4>},
|
|
|
|
{0,0,0,0}, {0,0,0,0},
|
|
|
|
{imgproc::upsampleCaller<short,1>, imgproc::upsampleCaller<short,2>,
|
|
|
|
imgproc::upsampleCaller<short,3>, imgproc::upsampleCaller<short,4>},
|
|
|
|
{0,0,0,0},
|
|
|
|
{imgproc::upsampleCaller<float,1>, imgproc::upsampleCaller<float,2>,
|
|
|
|
imgproc::upsampleCaller<float,3>, imgproc::upsampleCaller<float,4>}};
|
|
|
|
|
|
|
|
Caller caller = callers[src.depth()][src.channels()-1];
|
|
|
|
if (!caller)
|
|
|
|
CV_Error(CV_StsUnsupportedFormat, "bad number of channels");
|
|
|
|
|
|
|
|
dst.create(src.rows*2, src.cols*2, src.type());
|
2011-08-01 16:15:31 +08:00
|
|
|
caller(src, dst.reshape(1), StreamAccessor::getStream(stream));
|
2011-06-30 22:39:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// pyrDown
|
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
2011-06-30 22:39:48 +08:00
|
|
|
{
|
2011-08-31 19:42:54 +08:00
|
|
|
template <typename T, int cn> void pyrDown_gpu(const DevMem2D& src, const DevMem2D& dst, int borderType, cudaStream_t stream);
|
2011-08-24 19:16:42 +08:00
|
|
|
}}}
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
void cv::gpu::pyrDown(const GpuMat& src, GpuMat& dst, int borderType, Stream& stream)
|
2011-08-01 16:15:31 +08:00
|
|
|
{
|
2011-08-24 19:16:42 +08:00
|
|
|
using namespace cv::gpu::imgproc;
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
typedef void (*func_t)(const DevMem2D& src, const DevMem2D& dst, int borderType, cudaStream_t stream);
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
static const func_t funcs[6][4] =
|
2011-08-01 16:15:31 +08:00
|
|
|
{
|
2011-08-24 19:16:42 +08:00
|
|
|
{pyrDown_gpu<uchar, 1>, pyrDown_gpu<uchar, 2>, pyrDown_gpu<uchar, 3>, pyrDown_gpu<uchar, 4>},
|
|
|
|
{pyrDown_gpu<schar, 1>, pyrDown_gpu<schar, 2>, pyrDown_gpu<schar, 3>, pyrDown_gpu<schar, 4>},
|
|
|
|
{pyrDown_gpu<ushort, 1>, pyrDown_gpu<ushort, 2>, pyrDown_gpu<ushort, 3>, pyrDown_gpu<ushort, 4>},
|
|
|
|
{pyrDown_gpu<short, 1>, pyrDown_gpu<short, 2>, pyrDown_gpu<short, 3>, pyrDown_gpu<short, 4>},
|
|
|
|
{pyrDown_gpu<int, 1>, pyrDown_gpu<int, 2>, pyrDown_gpu<int, 3>, pyrDown_gpu<int, 4>},
|
|
|
|
{pyrDown_gpu<float, 1>, pyrDown_gpu<float, 2>, pyrDown_gpu<float, 3>, pyrDown_gpu<float, 4>},
|
|
|
|
};
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
|
|
|
|
|
2011-08-31 21:21:14 +08:00
|
|
|
CV_Assert(borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP);
|
2011-08-31 19:42:54 +08:00
|
|
|
int gpuBorderType;
|
|
|
|
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
|
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
dst.create((src.rows + 1) / 2, (src.cols + 1) / 2, src.type());
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
funcs[src.depth()][src.channels() - 1](src, dst, gpuBorderType, StreamAccessor::getStream(stream));
|
2011-06-30 22:39:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// pyrUp
|
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
namespace cv { namespace gpu { namespace imgproc
|
2011-08-01 16:15:31 +08:00
|
|
|
{
|
2011-08-31 19:42:54 +08:00
|
|
|
template <typename T, int cn> void pyrUp_gpu(const DevMem2D& src, const DevMem2D& dst, int borderType, cudaStream_t stream);
|
2011-08-24 19:16:42 +08:00
|
|
|
}}}
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
void cv::gpu::pyrUp(const GpuMat& src, GpuMat& dst, int borderType, Stream& stream)
|
2011-08-01 16:15:31 +08:00
|
|
|
{
|
2011-08-24 19:16:42 +08:00
|
|
|
using namespace cv::gpu::imgproc;
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
typedef void (*func_t)(const DevMem2D& src, const DevMem2D& dst, int borderType, cudaStream_t stream);
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
static const func_t funcs[6][4] =
|
2011-08-01 16:15:31 +08:00
|
|
|
{
|
2011-08-24 19:16:42 +08:00
|
|
|
{pyrUp_gpu<uchar, 1>, pyrUp_gpu<uchar, 2>, pyrUp_gpu<uchar, 3>, pyrUp_gpu<uchar, 4>},
|
|
|
|
{pyrUp_gpu<schar, 1>, pyrUp_gpu<schar, 2>, pyrUp_gpu<schar, 3>, pyrUp_gpu<schar, 4>},
|
|
|
|
{pyrUp_gpu<ushort, 1>, pyrUp_gpu<ushort, 2>, pyrUp_gpu<ushort, 3>, pyrUp_gpu<ushort, 4>},
|
|
|
|
{pyrUp_gpu<short, 1>, pyrUp_gpu<short, 2>, pyrUp_gpu<short, 3>, pyrUp_gpu<short, 4>},
|
|
|
|
{pyrUp_gpu<int, 1>, pyrUp_gpu<int, 2>, pyrUp_gpu<int, 3>, pyrUp_gpu<int, 4>},
|
|
|
|
{pyrUp_gpu<float, 1>, pyrUp_gpu<float, 2>, pyrUp_gpu<float, 3>, pyrUp_gpu<float, 4>},
|
|
|
|
};
|
2011-08-01 16:15:31 +08:00
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
|
|
|
|
|
2011-08-31 21:21:14 +08:00
|
|
|
CV_Assert(borderType == BORDER_REFLECT101 || borderType == BORDER_REPLICATE || borderType == BORDER_CONSTANT || borderType == BORDER_REFLECT || borderType == BORDER_WRAP);
|
2011-08-31 19:42:54 +08:00
|
|
|
int gpuBorderType;
|
|
|
|
CV_Assert(tryConvertToGpuBorderType(borderType, gpuBorderType));
|
|
|
|
|
2011-08-24 19:16:42 +08:00
|
|
|
dst.create(src.rows*2, src.cols*2, src.type());
|
|
|
|
|
2011-08-31 19:42:54 +08:00
|
|
|
funcs[src.depth()][src.channels() - 1](src, dst, gpuBorderType, StreamAccessor::getStream(stream));
|
2011-04-08 16:04:56 +08:00
|
|
|
}
|
2010-11-30 16:04:37 +08:00
|
|
|
|
2011-08-08 16:53:55 +08:00
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////////
|
|
|
|
// Canny
|
|
|
|
|
|
|
|
cv::gpu::CannyBuf::CannyBuf(const GpuMat& dx_, const GpuMat& dy_) : dx(dx_), dy(dy_)
|
|
|
|
{
|
|
|
|
CV_Assert(dx_.type() == CV_32SC1 && dy_.type() == CV_32SC1 && dx_.size() == dy_.size());
|
|
|
|
|
|
|
|
create(dx_.size(), -1);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::CannyBuf::create(const Size& image_size, int apperture_size)
|
|
|
|
{
|
|
|
|
ensureSizeIsEnough(image_size, CV_32SC1, dx);
|
|
|
|
ensureSizeIsEnough(image_size, CV_32SC1, dy);
|
|
|
|
|
|
|
|
if (apperture_size == 3)
|
|
|
|
{
|
|
|
|
ensureSizeIsEnough(image_size, CV_32SC1, dx_buf);
|
|
|
|
ensureSizeIsEnough(image_size, CV_32SC1, dy_buf);
|
|
|
|
}
|
|
|
|
else if(apperture_size > 0)
|
|
|
|
{
|
|
|
|
if (!filterDX)
|
|
|
|
filterDX = createDerivFilter_GPU(CV_8UC1, CV_32S, 1, 0, apperture_size, BORDER_REPLICATE);
|
|
|
|
if (!filterDY)
|
|
|
|
filterDY = createDerivFilter_GPU(CV_8UC1, CV_32S, 0, 1, apperture_size, BORDER_REPLICATE);
|
|
|
|
}
|
|
|
|
|
|
|
|
ensureSizeIsEnough(image_size.height + 2, image_size.width + 2, CV_32FC1, edgeBuf);
|
|
|
|
|
|
|
|
ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf1);
|
|
|
|
ensureSizeIsEnough(1, image_size.width * image_size.height, CV_16UC2, trackBuf2);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::CannyBuf::release()
|
|
|
|
{
|
|
|
|
dx.release();
|
|
|
|
dy.release();
|
|
|
|
dx_buf.release();
|
|
|
|
dy_buf.release();
|
|
|
|
edgeBuf.release();
|
|
|
|
trackBuf1.release();
|
|
|
|
trackBuf2.release();
|
|
|
|
}
|
|
|
|
|
|
|
|
namespace cv { namespace gpu { namespace canny
|
|
|
|
{
|
|
|
|
void calcSobelRowPass_gpu(PtrStep src, PtrStepi dx_buf, PtrStepi dy_buf, int rows, int cols);
|
|
|
|
|
|
|
|
void calcMagnitude_gpu(PtrStepi dx_buf, PtrStepi dy_buf, PtrStepi dx, PtrStepi dy, PtrStepf mag, int rows, int cols, bool L2Grad);
|
|
|
|
void calcMagnitude_gpu(PtrStepi dx, PtrStepi dy, PtrStepf mag, int rows, int cols, bool L2Grad);
|
|
|
|
|
|
|
|
void calcMap_gpu(PtrStepi dx, PtrStepi dy, PtrStepf mag, PtrStepi map, int rows, int cols, float low_thresh, float high_thresh);
|
|
|
|
|
|
|
|
void edgesHysteresisLocal_gpu(PtrStepi map, ushort2* st1, int rows, int cols);
|
|
|
|
|
|
|
|
void edgesHysteresisGlobal_gpu(PtrStepi map, ushort2* st1, ushort2* st2, int rows, int cols);
|
|
|
|
|
|
|
|
void getEdges_gpu(PtrStepi map, PtrStep dst, int rows, int cols);
|
|
|
|
}}}
|
|
|
|
|
|
|
|
namespace
|
|
|
|
{
|
|
|
|
void CannyCaller(CannyBuf& buf, GpuMat& dst, float low_thresh, float high_thresh)
|
|
|
|
{
|
|
|
|
using namespace cv::gpu::canny;
|
|
|
|
|
|
|
|
calcMap_gpu(buf.dx, buf.dy, buf.edgeBuf, buf.edgeBuf, dst.rows, dst.cols, low_thresh, high_thresh);
|
|
|
|
|
|
|
|
edgesHysteresisLocal_gpu(buf.edgeBuf, buf.trackBuf1.ptr<ushort2>(), dst.rows, dst.cols);
|
|
|
|
|
|
|
|
edgesHysteresisGlobal_gpu(buf.edgeBuf, buf.trackBuf1.ptr<ushort2>(), buf.trackBuf2.ptr<ushort2>(), dst.rows, dst.cols);
|
|
|
|
|
|
|
|
getEdges_gpu(buf.edgeBuf, dst, dst.rows, dst.cols);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::Canny(const GpuMat& src, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
|
|
|
|
{
|
|
|
|
CannyBuf buf(src.size(), apperture_size);
|
|
|
|
Canny(src, buf, dst, low_thresh, high_thresh, apperture_size, L2gradient);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::Canny(const GpuMat& src, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, int apperture_size, bool L2gradient)
|
|
|
|
{
|
|
|
|
using namespace cv::gpu::canny;
|
|
|
|
|
2011-09-05 14:20:34 +08:00
|
|
|
CV_Assert(TargetArchs::builtWith(SHARED_ATOMICS) && DeviceInfo().supports(SHARED_ATOMICS));
|
2011-08-08 16:53:55 +08:00
|
|
|
CV_Assert(src.type() == CV_8UC1);
|
|
|
|
|
|
|
|
if( low_thresh > high_thresh )
|
|
|
|
std::swap( low_thresh, high_thresh);
|
|
|
|
|
|
|
|
dst.create(src.size(), CV_8U);
|
|
|
|
dst.setTo(Scalar::all(0));
|
|
|
|
|
|
|
|
buf.create(src.size(), apperture_size);
|
|
|
|
buf.edgeBuf.setTo(Scalar::all(0));
|
|
|
|
|
|
|
|
if (apperture_size == 3)
|
|
|
|
{
|
|
|
|
calcSobelRowPass_gpu(src, buf.dx_buf, buf.dy_buf, src.rows, src.cols);
|
|
|
|
|
|
|
|
calcMagnitude_gpu(buf.dx_buf, buf.dy_buf, buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
buf.filterDX->apply(src, buf.dx, Rect(0, 0, src.cols, src.rows));
|
|
|
|
buf.filterDY->apply(src, buf.dy, Rect(0, 0, src.cols, src.rows));
|
|
|
|
|
|
|
|
calcMagnitude_gpu(buf.dx, buf.dy, buf.edgeBuf, src.rows, src.cols, L2gradient);
|
|
|
|
}
|
|
|
|
|
|
|
|
CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
|
|
|
|
{
|
|
|
|
CannyBuf buf(dx, dy);
|
|
|
|
Canny(dx, dy, buf, dst, low_thresh, high_thresh, L2gradient);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::Canny(const GpuMat& dx, const GpuMat& dy, CannyBuf& buf, GpuMat& dst, double low_thresh, double high_thresh, bool L2gradient)
|
|
|
|
{
|
|
|
|
using namespace cv::gpu::canny;
|
|
|
|
|
2011-09-05 14:20:34 +08:00
|
|
|
CV_Assert(TargetArchs::builtWith(SHARED_ATOMICS) && DeviceInfo().supports(SHARED_ATOMICS));
|
2011-08-08 16:53:55 +08:00
|
|
|
CV_Assert(dx.type() == CV_32SC1 && dy.type() == CV_32SC1 && dx.size() == dy.size());
|
|
|
|
|
|
|
|
if( low_thresh > high_thresh )
|
|
|
|
std::swap( low_thresh, high_thresh);
|
|
|
|
|
|
|
|
dst.create(dx.size(), CV_8U);
|
|
|
|
dst.setTo(Scalar::all(0));
|
|
|
|
|
|
|
|
buf.dx = dx; buf.dy = dy;
|
|
|
|
buf.create(dx.size(), -1);
|
|
|
|
buf.edgeBuf.setTo(Scalar::all(0));
|
|
|
|
|
|
|
|
calcMagnitude_gpu(dx, dy, buf.edgeBuf, dx.rows, dx.cols, L2gradient);
|
|
|
|
|
|
|
|
CannyCaller(buf, dst, static_cast<float>(low_thresh), static_cast<float>(high_thresh));
|
|
|
|
}
|
|
|
|
|
2010-08-31 15:23:03 +08:00
|
|
|
#endif /* !defined (HAVE_CUDA) */
|
2010-12-02 17:07:13 +08:00
|
|
|
|
2010-12-22 21:46:06 +08:00
|
|
|
|