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

Merge pull request #2618 from vbystricky:ipp_Scharr

Browse Source
This commit is contained in:
Alexander Alekhin 2014-04-22 15:02:12 +04:00 committed by OpenCV Buildbot
commit 580559e1e6
1 changed files with 260 additions and 130 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

390
modules/imgproc/src/deriv.cpp
View File

@ -185,11 +185,141 @@ cv::Ptr<cv::FilterEngine> cv::createDerivFilter(int srcType, int dstType,
#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7) #if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
#define IPP_RETURN_ERROR {setIppErrorStatus(); return false;}
namespace cv namespace cv
{ {
#if (IPP_VERSION_X100 >= 801)
static bool IPPDerivScharr(const Mat& src, Mat& dst, int ddepth, int dx, int dy, double scale) static bool IPPDerivScharr(InputArray _src, OutputArray _dst, int ddepth, int dx, int dy, double scale, double delta, int borderType)
{ {
#if defined(HAVE_IPP_ICV_ONLY)
_src; _dst; ddepth; dx; dy; scale; delta; borderType;
return false;
#else
if ((0 > dx) || (0 > dy) || (1 != dx + dy))
return false;
if (fabs(delta) > FLT_EPSILON)
return false;
IppiBorderType ippiBorderType = ippiGetBorderType(borderType & (~BORDER_ISOLATED));
if ((int)ippiBorderType < 0)
return false;
int stype = _src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype);
if (ddepth < 0)
ddepth = sdepth;
int dtype = CV_MAKETYPE(ddepth, cn);
Mat src = _src.getMat();
if (0 == (BORDER_ISOLATED & borderType))
{
Size size; Point offset;
src.locateROI(size, offset);
if (0 < offset.x)
ippiBorderType = (IppiBorderType)(ippiBorderType | ippBorderInMemLeft);
if (0 < offset.y)
ippiBorderType = (IppiBorderType)(ippiBorderType | ippBorderInMemTop);
if (offset.x + src.cols < size.width)
ippiBorderType = (IppiBorderType)(ippiBorderType | ippBorderInMemRight);
if (offset.y + src.rows < size.height)
ippiBorderType = (IppiBorderType)(ippiBorderType | ippBorderInMemBottom);
}
bool horz = (0 == dx) && (1 == dy);
IppiSize roiSize = {src.cols, src.rows};
_dst.create( _src.size(), dtype);
Mat dst = _dst.getMat();
IppStatus sts = ippStsErr;
if ((CV_8U == stype) && (CV_16S == dtype))
{
int bufferSize = 0; Ipp8u *pBuffer;
if (horz)
{
if (0 > ippiFilterScharrHorizMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp8u, ipp16s, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrHorizMaskBorder_8u16s_C1R(src.data, (int)src.step, (Ipp16s *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
else
{
if (0 > ippiFilterScharrVertMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp8u, ipp16s, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrVertMaskBorder_8u16s_C1R(src.data, (int)src.step, (Ipp16s *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
ippsFree(pBuffer);
}
else if ((CV_16S == stype) && (CV_16S == dtype))
{
int bufferSize = 0; Ipp8u *pBuffer;
if (horz)
{
if (0 > ippiFilterScharrHorizMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp16s, ipp16s, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrHorizMaskBorder_16s_C1R((Ipp16s *)src.data, (int)src.step, (Ipp16s *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
else
{
if (0 > ippiFilterScharrVertMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp16s, ipp16s, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrVertMaskBorder_16s_C1R((Ipp16s *)src.data, (int)src.step, (Ipp16s *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
ippsFree(pBuffer);
}
else if ((CV_32F == stype) && (CV_32F == dtype))
{
int bufferSize = 0; Ipp8u *pBuffer;
if (horz)
{
if (0 > ippiFilterScharrHorizMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp32f, ipp32f, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrHorizMaskBorder_32f_C1R((Ipp32f *)src.data, (int)src.step, (Ipp32f *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
else
{
if (0 > ippiFilterScharrVertMaskBorderGetBufferSize(roiSize, ippMskSize3x3, ipp32f, ipp32f, 1, &bufferSize))
IPP_RETURN_ERROR
pBuffer = ippsMalloc_8u(bufferSize);
if (NULL == pBuffer)
IPP_RETURN_ERROR
sts = ippiFilterScharrVertMaskBorder_32f_C1R((Ipp32f *)src.data, (int)src.step, (Ipp32f *)dst.data, (int)dst.step, roiSize, ippMskSize3x3, ippiBorderType, 0, pBuffer);
}
ippsFree(pBuffer);
if (sts < 0)
IPP_RETURN_ERROR;
if (FLT_EPSILON < fabs(scale - 1.0))
sts = ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, roiSize);
}
return (0 <= sts);
#endif
}
#elif (IPP_VERSION_MAJOR >= 7)
static bool IPPDerivScharr(InputArray _src, OutputArray _dst, int ddepth, int dx, int dy, double scale, double delta, int borderType)
{
if (BORDER_REPLICATE != borderType)
return false;
if ((0 > dx) || (0 > dy) || (1 != dx + dy))
return false;
if (fabs(delta) > FLT_EPSILON)
return false;
Mat src = _src.getMat(), dst = _dst.getMat();
int bufSize = 0; int bufSize = 0;
cv::AutoBuffer<char> buffer; cv::AutoBuffer<char> buffer;
IppiSize roi = ippiSize(src.cols, src.rows); IppiSize roi = ippiSize(src.cols, src.rows);
@ -285,142 +415,145 @@ static bool IPPDerivScharr(const Mat& src, Mat& dst, int ddepth, int dx, int dy,
return false; return false;
} }
} }
#endif
static bool IPPDerivSobel(InputArray _src, OutputArray _dst, int ddepth, int dx, int dy, int ksize, double scale, double delta, int borderType)
static bool IPPDeriv(const Mat& src, Mat& dst, int ddepth, int dx, int dy, int ksize, double scale)
{ {
int stype = src.type(), dtype = dst.type(), bufSize = 0; if ((borderType != BORDER_REPLICATE) || (3 != ksize) || (5 != ksize))
return false;
if (fabs(delta) > FLT_EPSILON)
return false;
if (1 != _src.channels())
return false;
int bufSize = 0;
cv::AutoBuffer<char> buffer; cv::AutoBuffer<char> buffer;
if (ksize == 3 || ksize == 5) Mat src = _src.getMat(), dst = _dst.getMat();
if ( ddepth < 0 )
ddepth = src.depth();
if (src.type() == CV_8U && dst.type() == CV_16S && scale == 1)
{ {
if (stype == CV_8UC1 && dtype == CV_16SC1 && scale == 1) if ((dx == 1) && (dy == 0))
{ {
if (dx == 1 && dy == 0) if (0 > ippiFilterSobelNegVertGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
{ IPP_RETURN_ERROR
if (0 > ippiFilterSobelNegVertGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize)) buffer.allocate(bufSize);
return false;
buffer.allocate(bufSize);
return 0 <= ippiFilterSobelNegVertBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step, if (0 > ippiFilterSobelNegVertBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), (Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer); ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
} IPP_RETURN_ERROR
return true;
if (dx == 0 && dy == 1)
{
if (0 > ippiFilterSobelHorizGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false;
buffer.allocate(bufSize);
return 0 <= ippiFilterSobelHorizBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer);
}
if (dx == 2 && dy == 0)
{
if (0 > ippiFilterSobelVertSecondGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false;
buffer.allocate(bufSize);
return 0 <= ippiFilterSobelVertSecondBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer);
}
if (dx == 0 && dy == 2)
{
if (0 > ippiFilterSobelHorizSecondGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false;
buffer.allocate(bufSize);
return 0 <= ippiFilterSobelHorizSecondBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer);
}
} }
if (stype == CV_32FC1 && dtype == CV_32FC1) if ((dx == 0) && (dy == 1))
{ {
#if defined(HAVE_IPP_ICV_ONLY) // N/A: ippiMulC_32f_C1R if (0 > ippiFilterSobelHorizGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false; IPP_RETURN_ERROR
#else buffer.allocate(bufSize);
#if 0
if (dx == 1 && dy == 0)
{
if (0 > ippiFilterSobelNegVertGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), &bufSize))
return false;
buffer.allocate(bufSize);
if (0 > ippiFilterSobelNegVertBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step, if (0 > ippiFilterSobelHorizBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), (Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer)) ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
{ IPP_RETURN_ERROR
return false; return true;
} }
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
if (dx == 0 && dy == 1) if ((dx == 2) && (dy == 0))
{ {
if (0 > ippiFilterSobelHorizGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize)) if (0 > ippiFilterSobelVertSecondGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false; IPP_RETURN_ERROR
buffer.allocate(bufSize); buffer.allocate(bufSize);
if (0 > ippiFilterSobelHorizBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step, if (0 > ippiFilterSobelVertSecondBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), (Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer)) ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
{ IPP_RETURN_ERROR
return false; return true;
} }
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
#endif
if(dx == 2 && dy == 0) if ((dx == 0) && (dy == 2))
{ {
if (0 > ippiFilterSobelVertSecondGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize)) if (0 > ippiFilterSobelHorizSecondGetBufferSize_8u16s_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false; IPP_RETURN_ERROR
buffer.allocate(bufSize); buffer.allocate(bufSize);
if (0 > ippiFilterSobelVertSecondBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step, if (0 > ippiFilterSobelHorizSecondBorder_8u16s_C1R((const Ipp8u*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), (Ipp16s*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer)) ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
{ IPP_RETURN_ERROR
return false; return true;
}
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
if(dx == 0 && dy == 2)
{
if (0 > ippiFilterSobelHorizSecondGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
return false;
buffer.allocate(bufSize);
if (0 > ippiFilterSobelHorizSecondBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
{
return false;
}
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
#endif
} }
} }
if (ksize <= 0) if (src.type() == CV_32F && dst.type() == CV_32F)
return IPPDerivScharr(src, dst, ddepth, dx, dy, scale); {
#if defined(HAVE_IPP_ICV_ONLY) // N/A: ippiMulC_32f_C1R
return false;
#else
#if 0
if ((dx == 1) && (dy == 0))
{
if (0 > ippiFilterSobelNegVertGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize), &bufSize))
IPP_RETURN_ERROR
buffer.allocate(bufSize);
if (0 > ippiFilterSobelNegVertBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
IPP_RETURN_ERROR
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
if ((dx == 0) && (dy == 1))
{
if (0 > ippiFilterSobelHorizGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
IPP_RETURN_ERROR
buffer.allocate(bufSize);
if (0 > ippiFilterSobelHorizBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
IPP_RETURN_ERROR
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
#endif
if((dx == 2) && (dy == 0))
{
if (0 > ippiFilterSobelVertSecondGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
IPP_RETURN_ERROR
buffer.allocate(bufSize);
if (0 > ippiFilterSobelVertSecondBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
IPP_RETURN_ERROR
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
if((dx == 0) && (dy == 2))
{
if (0 > ippiFilterSobelHorizSecondGetBufferSize_32f_C1R(ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),&bufSize))
IPP_RETURN_ERROR
buffer.allocate(bufSize);
if (0 > ippiFilterSobelHorizSecondBorder_32f_C1R((const Ipp32f*)src.data, (int)src.step,
(Ipp32f*)dst.data, (int)dst.step, ippiSize(src.cols, src.rows), (IppiMaskSize)(ksize*10+ksize),
ippBorderRepl, 0, (Ipp8u*)(char*)buffer))
IPP_RETURN_ERROR
if(scale != 1)
ippiMulC_32f_C1R((Ipp32f *)dst.data, (int)dst.step, (Ipp32f)scale, (Ipp32f *)dst.data, (int)dst.step, ippiSize(dst.cols*dst.channels(), dst.rows));
return true;
}
#endif
}
return false; return false;
} }
@ -449,11 +582,14 @@ void cv::Sobel( InputArray _src, OutputArray _dst, int ddepth, int dx, int dy,
#endif #endif
#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7) #if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
if (dx < 3 && dy < 3 && cn == 1 && borderType == BORDER_REPLICATE && if (ksize < 0)
((stype == CV_8UC1 && dtype == CV_16SC1) || (stype == CV_32FC1 && dtype == CV_32FC1)))
{ {
Mat src = _src.getMat(), dst = _dst.getMat(); if (IPPDerivScharr(_src, _dst, ddepth, dx, dy, scale, delta, borderType))
if (IPPDeriv(src, dst, ddepth, dx, dy, ksize,scale)) return;
}
else if (0 < ksize)
{
if (IPPDerivSobel(_src, _dst, ddepth, dx, dy, ksize, scale, delta, borderType))
return; return;
} }
#endif #endif
@ -493,14 +629,8 @@ void cv::Scharr( InputArray _src, OutputArray _dst, int ddepth, int dx, int dy,
#endif #endif
#if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7) #if defined (HAVE_IPP) && (IPP_VERSION_MAJOR >= 7)
if(dx < 2 && dy < 2 && borderType == BORDER_REPLICATE && if (IPPDerivScharr(_src, _dst, ddepth, dx, dy, scale, delta, borderType))
((stype == CV_8UC1 && dtype == CV_16SC1) || (stype == CV_32FC1 && dtype == CV_32FC1))) return;
{
Mat src = _src.getMat(), dst = _dst.getMat();
if(IPPDerivScharr(src, dst, ddepth, dx, dy, scale))
return;
setIppErrorStatus();
}
#endif #endif
int ktype = std::max(CV_32F, std::max(ddepth, sdepth)); int ktype = std::max(CV_32F, std::max(ddepth, sdepth));