How can we use DMIP Kernels in functional level implementation? For example in functional level implementation of sobel operation we require kernel to get multiplied with the source image. When we are trying to link the kernel with the source image it is not happening since it is not a base node type. Is it the right way or we should implement only with the symbolic level implementation?
链接已复制
3 回复数
Hi,
If you refer to article on Deferred Mode Image Processing Framework, it shows how to implement IPP based Sobel edge detector code and how top implement same using DMIP.
Also in the 2nd code snippet(added below), you can see how to add source image in to DMIP and link with kernel.
[bash]1. Image Src(pSrcImg, Ipp8u, IppC3); // Source image in DMIP format 2. Image Dst(pDstImg, Ipp8u, IppC1); // Destination image in DMIP format 3. 4. Kernel Kh(idmFilterSobelHoriz,ippMskSize3x3,ipp8u,ipp16s); // Dx operator 5. Kernel Kv(idmFilterSobelVert,ippMskSize3x3,ipp8u,ipp16s); // Dy operator 6. 7. Graph O = ToGray(Src); // To get detected as common expression. 8. 9. // Compile end execute task 10. Dst = To8u(Abs(O*Kh)+ Abs(O*Kv)); [/bash]
Regards,
Naveen Gv
Naveen,
Above code is working fine.
We also implemented the same using DMIP functional level. We used FilterNode and initialized this node with Kernels.
// Define Horizantal and vertical Kernels
Kernel KH(idmFilterSobelHoriz);
Kernel KV(idmFilterSobelVert);
// Initialize Filter nodes with respective kernels
FilterNode *nd1=new FilterNode();
nd1->Init(&KH);
FilterNode *nd2=new FilterNode();
nd2->Init(&KV);
Regards,
Kishor.
Above code is working fine.
We also implemented the same using DMIP functional level. We used FilterNode and initialized this node with Kernels.
// Define Horizantal and vertical Kernels
Kernel KH(idmFilterSobelHoriz);
Kernel KV(idmFilterSobelVert);
// Initialize Filter nodes with respective kernels
FilterNode *nd1=new FilterNode();
nd1->Init(&KH);
FilterNode *nd2=new FilterNode();
nd2->Init(&KV);
Regards,
Kishor.