topic IPP erode issue when eroding entire image in Intel® Integrated Performance Primitives

IPP erode issue when eroding entire image

cks2k2 — Wed, 17 Aug 2011 03:33:30 GMT

I am trying to apply an erode operation to an entire grayscale image.

Realizing the issues with borders, I use ippiErodeBorderReplicate() but a cross-check with the expected output shows that some pixels (~10) at the top and bottom border are incorrect.

[bash]Ipp8u mask[9] = {0,1,0,1,1,1,0,1,0}; // some random kernel
IppiSize maskSize = {3,3};
IppiPoint anchor = {1,1};
IppiMorphState* morphState = NULL;
IppStatus status;

ippiMorphologyInitAlloc_8u_C1R(dstRoiSize.width, mask, maskSize, anchor, &morphState);

status = ippiErodeBorderReplicate_8u_C1R(pSrc, srcStep, pDst, dstStep, dstRoiSize, ippBorderRepl, morphState);

ippiMorphologyFree(morphState);[/bash]

A copy of the source image is created before the code above is called; source and image have exact same dimensions. ROI has same dimension as source and destination.

IPP erode issue when eroding entire image

Ying_H_Intel — Wed, 17 Aug 2011 08:50:11 GMT

Hello cks2k2, Coudl you please provide the input and output, And if possible, please add the below code // Print the version of ipp being used const IppLibraryVersion* lib = ippiGetLibVersion(); printf("%s %s %d.%d.%d.%d\n", lib->Name, lib->Version,lib->major, lib->minor, lib->majorBuild, lib->build); in your code and and let me know the result? Regards, Ying H.

IPP erode issue when eroding entire image

cks2k2 — Thu, 18 Aug 2011 05:55:16 GMT

Hi the output that you requested:

lib name: ippiv8l.lib

version: 5.2.108.410

On IPP 5.2.

IPP erode issue when eroding entire image

Ying_H_Intel — Thu, 18 Aug 2011 08:54:50 GMT

Have you try the latest version?
( geta 30-dayseval version fromIntel Downloads, Registration and Licensing

As you haven't provide the test image,Itried the sample 8-2

#include
#include
#include "ipp.h"

int main()
{

// Print the version of ipp being used
const IppLibraryVersion* lib = ippiGetLibVersion();
printf("%s %s %d.%d.%d.%d\n", lib->Name, lib->Version,lib->major, lib->minor, lib->majorBuild, lib->build);

IppiMorphState* ppState;
IppiSize roiSize = {5, 5};
Ipp8u pMask[3*3] = {1, 1, 1,1, 0, 1,1, 1, 1};
IppiSize maskSize = {3, 3};
IppiPoint anchor = {1, 1};
Ipp8u pSrc[5*5] = {
1, 2, 4, 1, 2,
5, 1, 2, 1, 2,
1, 2, 1, 2, 1,
2, 1, 5, 1, 2};
int srcStep = 5;
int dstStep = 5;
Ipp8u pDst[5*5];
int dstSize = 5;
IppStatus status;
ippiMorphologyInitAlloc_8u_C1R( roiSize.width, pMask, maskSize, anchor,
&ppState);
status = ippiErodeBorderReplicate_8u_C1R( pSrc, srcStep, pDst, dstStep, roiSize,
ippBorderRepl, ppState);

printf ("ippiErodeBorderReplicate return %s\n", ippGetStatusString(status));
ippiMorphologyFree(ppState);
int i=0, j=0;
for ( i=0; i<5; i++)
{
for (j=0;j<5; j++)
printf("%d \t", pDst[i*5+j]);
printf("\n");
}

return 0;

}
I have the below result.

ippiv8-7.0.dll 7.0 build 205.58 7.0.205.1054
ippiErodeBorderReplicate return ippStsNoErr: No error, it's OK
1 1 1 1 1
1 1 1 1 1
1 1 1 1 1
0 0 0 0 0
0 0 0 0 0
Press any key to continue . . .

Please try it and let me know if any result.

Regards,
Ying H.

IPP erode issue when eroding entire image

cks2k2 — Fri, 19 Aug 2011 02:10:37 GMT

I have attached the test image. Kernel is 3x3, with values {1,1,1, 0,1,0, 0,1,0} and anchor {1,1}. (T-kernel).

IPP erode issue when eroding entire image

Ying_H_Intel — Fri, 19 Aug 2011 06:37:26 GMT

Hello Cks2k2,

Could you tell theplace of wrong result you are seeing? And if possible, please try the latest Ipp version?

I print some value of the test image and result image. Theirresultseems ok.

With T mask.

Src:

src[0] [0:9] and src [1][0:9]

103 95 95 102 108 112 112 109 107 108

110 103 104 110 113 113 112 108 108 110

Dst:
dst[0] [0:9] anddst [1][0:9]

95 95 95 95 102 108 109 107 107 107
95 95 95 95 102 108 109 107 107 107

src[510] [250:259] and src [511][250:259]

(110) 112 114 111 109 111 111 115 106 113 112

(110) 111 113 110 108 109 110 114 105 112 111

dst[510] [250:259] anddst [511][250:259]

111 112 110 108 109 110 107 105 107 111

110 111 109 108 109 110 106 105 106 111

Regards,
Ying