During migrating my codes from IPP7 into IPP8, I found that ippiResizeSqrPixel functions are deprecated.
I use them to resize medical data. Fine control of zoom factor and shift is crucial because we need them to overlay multiple medical images at exact position.
However, new ippiResize* functions don't take zoom factor / shift argument any more.
Only substitute for ippiResizeSqrPixel seem to be ippiWarpAffine function, but ippiWarpAffine is too slow compared with ippiResizeSqrPixel.
Is there any other way to resize image with zoom factor / shift?
I've exactly the same problem. I need to resize an image and I need to specify zoom factor and shift.
I can't specify implicitly zoom factor by input and output roi size. ippiResizeSqrPixel is the right solution
for my problem. Would be nice if this function will not be deprecated.
It is only this particular function declaration that has been marked as depreciated.
Its interpolation parameter has been removed and a new collection of function declarations has been added, each implementing one interpolation type.
It isn't only a problem of function declaration. The new ippiResizeXXX family has an interface like this :
IppStatus ippiResizeLinear_<mod>(const Ipp<datatype>* pSrc, Ipp32s srcStep, Ipp<datatype>* pDst, Ipp32s dstStep, IppiPoint dstOffset, IppiSize dstSize, IppiBorderType border, Ipp<datatype>* borderValue, IppiResizeSpec_32f* pSpec, Ipp8u*pBuffer);
where zoom factor and shift are not present. Zoom factor is specified by source and destination size.
This is the problem.
I understand that if you need sub-pixel precision, the new function types won't make it. I'm just a user like you, so I guess Intel should step in and clarify the situation...
Thank you for your feedbacks about ResizeSqrPixel and WarpAffine performance. We are working on warp functions to reach better performance. Please wait updates of warp functionality in future IPP versions.
Best regards, Valentin
Glad to hear that you are working on warp functions.
; I hope warp function is as fast as ResizeSqrPixel; at this time, WarpAffine is 10x slower than ResizeSqrPixel.
It would be better that new resize functions take arguments for zoom and shift in sub pixel accuracy.