ipprWarpAffine

[Hadar_P_]

Hi

Is there a special reason why ipprWarpAffine (and the other 3d functions as well) does not support signed short data? Is there a way to bypass it?

Thanks

Hadar

[Ying_H_Intel]

Hi Hadar, 

Right, it had only 8u, 16u, 32f in current version because they were asked when implemented.  If your data are 16s, you may have to convert to 32f  ippsConvert_16s32f, then call 32f function. 

Do you like the 16s as a new feature?  if yes, could you please share some background information about your project, data size etc. If they are private Information, you can send us private message.

Best Regards,

Ying  H.

Intel IPP Support 

P.S 

IppStatus ipprWarpAffine_8u_C1V(const Ipp8u* pSrc, IpprVolume srcVolume, int srcStep, int srcPlaneStep, IpprCuboid srcVoi, Ipp8u* pDst, int dstStep, int dstPlaneStep, IpprCuboid dstVOI, const double coeffs[3][4], int interpolation);

IppStatus ipprWarpAffine_16u_C1V(const Ipp16u* pSrc, IpprVolume srcVolume, int srcStep, int srcPlaneStep, IpprCuboid srcVoi, Ipp16u* pDst, int dstStep, int dstPlaneStep, IpprCuboid dstVOI, const double coeffs[3][4], int interpolation);

IppStatus ipprWarpAffine_32f_C1V(const Ipp32f* pSrc, IpprVolume srcVolume, int srcStep, int srcPlaneStep, IpprCuboid srcVoi, Ipp32f* pDst, int dstStep, int dstPlaneStep, IpprCuboid dstVOI, const double coeffs[3][4], int interpolation);

[Hadar_P_]

Hi Ying

Thanks for the prompt answer!

I am dealing with medical data (hundreds of slices of 512*512 signed shorts usually, but also much bigger), and I would like to cut a slab or plane of this data, in general direction, with interpolation. 

Copying the data to 32f is impossible due to the cost in memory usage and time consumption.

Best

Hadar

[Hadar_P_]

To make it clear, I would like to have something like matlab's interp3

[Ying_H_Intel]

Hi Hadar,

Get it. matlab's interp3, I had escalated your request to IPP product definition team for considering in next product plan. I will let you know if any updates about the feature. DPD200579391

Best Regards,

Ying

[Hadar_P_]

Ying

Thanks a lot for your response, please add to the request the possibility to have different (float) factors for each axis, like different millimeter per pixel in each direction

Thanks again and best regards

Hadar