generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Added OpenCL support for FastNlMeansDenoisingAbs

Browse Source
This commit is contained in:
Erik Karlsson 2015-03-01 22:21:36 +01:00
parent 9f7cac8c59
commit a9ff335a89
2 changed files with 14 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
modules/photo/src/denoising.cpp
View File

@ -51,7 +51,7 @@ void cv::fastNlMeansDenoising( InputArray _src, OutputArray _dst, float h,
Size src_size = _src.size(); Size src_size = _src.size();
CV_OCL_RUN(_src.dims() <= 2 && (_src.isUMat() || _dst.isUMat()) && CV_OCL_RUN(_src.dims() <= 2 && (_src.isUMat() || _dst.isUMat()) &&
src_size.width > 5 && src_size.height > 5, // low accuracy on small sizes src_size.width > 5 && src_size.height > 5, // low accuracy on small sizes
ocl_fastNlMeansDenoising(_src, _dst, h, templateWindowSize, searchWindowSize)) ocl_fastNlMeansDenoising(_src, _dst, h, templateWindowSize, searchWindowSize, false))
Mat src = _src.getMat(); Mat src = _src.getMat();
_dst.create(src_size, src.type()); _dst.create(src_size, src.type());
@ -88,6 +88,10 @@ void cv::fastNlMeansDenoisingAbs( InputArray _src, OutputArray _dst, float h,
int templateWindowSize, int searchWindowSize) int templateWindowSize, int searchWindowSize)
{ {
Size src_size = _src.size(); Size src_size = _src.size();
CV_OCL_RUN(_src.dims() <= 2 && (_src.isUMat() || _dst.isUMat()) &&
src_size.width > 5 && src_size.height > 5, // low accuracy on small sizes
ocl_fastNlMeansDenoising(_src, _dst, h, templateWindowSize, searchWindowSize, true))
Mat src = _src.getMat(); Mat src = _src.getMat();
_dst.create(src_size, src.type()); _dst.create(src_size, src.type());
Mat dst = _dst.getMat(); Mat dst = _dst.getMat();

18
modules/photo/src/fast_nlmeans_denoising_opencl.hpp
View File

@ -30,7 +30,7 @@ static int divUp(int a, int b)
template <typename FT> template <typename FT>
static bool ocl_calcAlmostDist2Weight(UMat & almostDist2Weight, int searchWindowSize, int templateWindowSize, FT h, int cn, static bool ocl_calcAlmostDist2Weight(UMat & almostDist2Weight, int searchWindowSize, int templateWindowSize, FT h, int cn,
int & almostTemplateWindowSizeSqBinShift) int & almostTemplateWindowSizeSqBinShift, bool abs)
{ {
const int maxEstimateSumValue = searchWindowSize * searchWindowSize * 255; const int maxEstimateSumValue = searchWindowSize * searchWindowSize * 255;
int fixedPointMult = std::numeric_limits<int>::max() / maxEstimateSumValue; int fixedPointMult = std::numeric_limits<int>::max() / maxEstimateSumValue;
@ -48,15 +48,15 @@ static bool ocl_calcAlmostDist2Weight(UMat & almostDist2Weight, int searchWindow
FT almostDist2ActualDistMultiplier = (FT)(1 << almostTemplateWindowSizeSqBinShift) / templateWindowSizeSq; FT almostDist2ActualDistMultiplier = (FT)(1 << almostTemplateWindowSizeSqBinShift) / templateWindowSizeSq;
const FT WEIGHT_THRESHOLD = 1e-3f; const FT WEIGHT_THRESHOLD = 1e-3f;
int maxDist = 255 * 255 * cn; int maxDist = abs ? 255 * cn : 255 * 255 * cn;
int almostMaxDist = (int)(maxDist / almostDist2ActualDistMultiplier + 1); int almostMaxDist = (int)(maxDist / almostDist2ActualDistMultiplier + 1);
FT den = 1.0f / (h * h * cn); FT den = 1.0f / (h * h * cn);
almostDist2Weight.create(1, almostMaxDist, CV_32SC1); almostDist2Weight.create(1, almostMaxDist, CV_32SC1);
ocl::Kernel k("calcAlmostDist2Weight", ocl::photo::nlmeans_oclsrc, ocl::Kernel k("calcAlmostDist2Weight", ocl::photo::nlmeans_oclsrc,
format("-D OP_CALC_WEIGHTS -D FT=%s%s", ocl::typeToStr(depth), format("-D OP_CALC_WEIGHTS -D FT=%s%s%s", ocl::typeToStr(depth),
doubleSupport ? " -D DOUBLE_SUPPORT" : "")); doubleSupport ? " -D DOUBLE_SUPPORT" : "", abs ? " -D ABS" : ""));
if (k.empty()) if (k.empty())
return false; return false;
@ -68,7 +68,7 @@ static bool ocl_calcAlmostDist2Weight(UMat & almostDist2Weight, int searchWindow
} }
static bool ocl_fastNlMeansDenoising(InputArray _src, OutputArray _dst, float h, static bool ocl_fastNlMeansDenoising(InputArray _src, OutputArray _dst, float h,
int templateWindowSize, int searchWindowSize) int templateWindowSize, int searchWindowSize, bool abs)
{ {
int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type); int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
int ctaSize = ocl::Device::getDefault().isIntel() ? CTA_SIZE_INTEL : CTA_SIZE_DEFAULT; int ctaSize = ocl::Device::getDefault().isIntel() ? CTA_SIZE_INTEL : CTA_SIZE_DEFAULT;
@ -89,21 +89,21 @@ static bool ocl_fastNlMeansDenoising(InputArray _src, OutputArray _dst, float h,
" -D sample_t=%s -D pixel_t=%s -D int_t=%s" " -D sample_t=%s -D pixel_t=%s -D int_t=%s"
" -D BLOCK_COLS=%d -D BLOCK_ROWS=%d" " -D BLOCK_COLS=%d -D BLOCK_ROWS=%d"
" -D CTA_SIZE=%d -D TEMPLATE_SIZE2=%d -D SEARCH_SIZE2=%d" " -D CTA_SIZE=%d -D TEMPLATE_SIZE2=%d -D SEARCH_SIZE2=%d"
" -D convert_int_t=%s -D cn=%d -D convert_pixel_t=%s", " -D convert_int_t=%s -D cn=%d -D convert_pixel_t=%s%s",
templateWindowSize, searchWindowSize, templateWindowSize, searchWindowSize,
ocl::typeToStr(depth), ocl::typeToStr(type), ocl::typeToStr(CV_32SC(cn)), ocl::typeToStr(depth), ocl::typeToStr(type), ocl::typeToStr(CV_32SC(cn)),
BLOCK_COLS, BLOCK_ROWS, BLOCK_COLS, BLOCK_ROWS,
ctaSize, templateWindowHalfWize, searchWindowHalfSize, ctaSize, templateWindowHalfWize, searchWindowHalfSize,
ocl::convertTypeStr(CV_8U, CV_32S, cn, cvt[0]), type == CV_8UC3 ? 4 : cn, ocl::convertTypeStr(CV_8U, CV_32S, cn, cvt[0]), type == CV_8UC3 ? 4 : cn,
ocl::convertTypeStr(CV_32S, CV_8U, cn, cvt[1])); ocl::convertTypeStr(CV_32S, CV_8U, cn, cvt[1]), abs ? " -D ABS" : "");
ocl::Kernel k("fastNlMeansDenoising", ocl::photo::nlmeans_oclsrc, opts); ocl::Kernel k("fastNlMeansDenoising", ocl::photo::nlmeans_oclsrc, opts);
if (k.empty()) if (k.empty())
return false; return false;
UMat almostDist2Weight; UMat almostDist2Weight;
if (!ocl_calcAlmostDist2Weight<float>(almostDist2Weight, searchWindowSize, templateWindowSize, h, cn, if (!ocl_calcAlmostDist2Weight<float>(almostDist2Weight, searchWindowSize, templateWindowSize,
almostTemplateWindowSizeSqBinShift)) h, cn, almostTemplateWindowSizeSqBinShift, abs))
return false; return false;
CV_Assert(almostTemplateWindowSizeSqBinShift >= 0); CV_Assert(almostTemplateWindowSizeSqBinShift >= 0);