Apply the filter to image columns.
Case 1: Operation on integer data
IppStatus ippiFilterColumnPipeline_<mod>(const Ipp<srcDatatype>** ppSrc, Ipp<dstDatatype>* pDst, int dstStep, IppiSize dstRoiSize, const Ipp<srcDatatype>* pKernel, int kernelSize, int divisor, Ipp8u* pBuffer);
Supported values for mod:
16s_C1R | 16s8u_C1R | 16s8s_C1R |
16s_C3R | 16s8u_C3R | 16s8s_C3R |
IppStatus ippiFilterColumnPipeline_Low_16s_C1R(const Ipp16s** ppSrc, Ipp16s* pDst, int dstStep, IppiSize dstRoiSize, const Ipp16s* pKernel, int kernelSize, int divisor, Ipp8u* pBuffer);
IppStatus ippiFilterColumnPipeline_Low_16s_C3R(const Ipp16s** ppSrc, Ipp16s* pDst, int dstStep, IppiSize dstRoiSize, const Ipp16s* pKernel, int kernelSize, int divisor, Ipp8u* pBuffer);
Case 2: Operation on floating-point data
IppStatus ippiFilterColumnPipeline_<mod>(const Ipp<datatype>** ppSrc, Ipp<datatype>* pDst, int dstStep, IppiSize dstRoiSize, const Ipp<datatype>* pKernel, int kernelSize, Ipp8u* pBuffer);
Supported values for mod:
32f_C1R |
32f_C3R |
ppSrc |
Double pointer to the source image ROI. |
pDst |
Pointer to the destination image ROI. |
dstStep |
Distance in bytes between starts of consecutive lines in the destination image. |
dstRoiSize |
Size of the destination ROI in pixels. |
pKernel |
Pointer to the strow kernel values. |
kernelSize |
Size of the kernel in pixels. |
divisor |
Value by which the computed result is divided (for operations on integer data only). |
pBuffer |
Pointer to the working buffer. |
The functions ippiFilterColumnPipeline and ippiFilterColumnPipeline_Low are declared in the ippcv.h file. The operates with ROI (see Regions of Interest in Intel IPP).
The function ippiFilterColumnPipeline_Low performs calculation exclusively with the 16s-data, and the input data must be in the range ensuring that the overflow does not occur during calculation and the result can be represented by a 32-bit integer number.
These functions apply the column filter of the separable convolution kernel to the source image pSrc. The filter coefficients are placed in the reversed order. For integer data:
Here j = 0, ... dstRoiSize.width-1, i=0,... dstRoiSize.height-1.
The size of the source image is
(dstRoiSize.height + kernelSize - 1) * dstRoiSize.width.
The functions requires the external buffer pBuffer, its size should be previously computed by the functions ippiFilterColumnPipelineGetBufferSize and ippiFilterColumnPipelineGetBufferSize_Low respectively.
These functions can be used to organize the separable convolution as a step of image processing pipeline (see Example “Separable Convolution 3x3 by One Row”).
ippStsNoErr |
Indicates no error. Any other value indicates an error. |
ippStsNullPtrErr |
Indicates an error condition if one of the specified pointers is NULL. |
ippStsSizeErr |
Indicates an error condition if roiSize has a field with a zero or negative value. |
ippStsStepErr |
Indicates an error condition if srcStep or dstStep is less than roiSize.width * <pixelSize> |
ippStsNotEvenStepErr |
Indicates an error condition if one of the step values is not divisible by 4 for floating-point images, or by 2 for short-integer images. |
ippStsBadArgErr |
Indicates an error condition if divisor is equal to 0. |
void Separable_3x3(const Ipp16s* pSrc, int srcStep, Ipp16s* pDst, int dstStep, IppiSize roiSize, Ipp16s* pKerX, Ipp16s* pKerY) { Ipp16s **get, *dst=pDst; const Ipp16s *src=pSrc; IppiSize roi; int todo=roiSize.height,sizeRow,sizeCol,bufLen; int mStep=(roiSize.width+7)&(~7),sStep=srcStep>>1,dStep=dstStep>>1; Ipp8u *pBufRow, *pBufCol; ippiFilterRowBorderPipelineGetBufferSize_Low_16s_C1R(roiSize,3,&sizeRow); ippiFilterColumnPipelineGetBufferSize_Low_16s_C1R(roiSize,3,&sizeCol); bufLen = mStep*3*sizeof(Ipp16s)+4*sizeof(Ipp16s*); pBufRow = ippsMalloc_8u(sizeRow); pBufCol = ippsMalloc_8u(sizeCol); get = (Ipp16s**)ippsMalloc_8u(bufLen); get[0]=get[1]=(Ipp16s*)(get+4); get[2]=get[1]+mStep; get[3]=get[2]+mStep; roi.width = roiSize.width; roi.height = 1; ippiFilterRowBorderPipeline_Low_16s_C1R(src, srcStep, get, roi, pKerX, 3, 1, ippBorderRepl, 0, 1, pBufRow); if (--todo) { get[2] = get[0]; } else { get[2] = get[0] + mStep; get[3] = get[2] + mStep; for (; todo>0; src+=sStep, dst+=dStep, todo--) { ippiFilterRowBorderPipeline_Low_16s_C1R(src, srcStep, get+2, roi, pKerX, 3, 1, ippBorderRepl, 0, 1, pBufRow); ippiFilterColumnPipeline_Low_16s_C1R(get, dst, dstStep, roi, pKerY, 3, 1, pBufCol); get[0] = get[1]; get[1] = get[2]; get[2] = get[3]; get[3] = get[0]; } } ippiFilterColumnPipeline_Low_16s_C1R(get, dst, dstStep, roi, pKerY, 3, 1, pBufCol); ippsFree(pBufRow); ippsFree(pBufCol); ippsFree(get); }
Copyright © 2000 - 2010, Intel Corporation. All rights reserved.