topic Hi Bruno, in Intel® Integrated Performance Primitives

Feature request ippiConvert_16u8u_C1RSfs

BMart1 — Wed, 08 Aug 2018 17:12:04 GMT

ippiConvert_16u8s_C1RSfs already exists.

could you submit this feature

Gennady_F_Intel — Thu, 09 Aug 2018 04:40:37 GMT

could you submit this feature request to the intel online service center?

Thanks, done.

BMart1 — Thu, 09 Aug 2018 18:30:06 GMT

Thanks, done.

Hi Bruno,

Igor_A_Intel — Fri, 10 Aug 2018 08:38:26 GMT

Hi Bruno,

you can use a workaround which is more flexible:

IPPAPI(IppStatus, ippiScaleC_16u8u_C1R , ( const Ipp16u* pSrc, int srcStep, Ipp64f mVal, Ipp64f aVal, Ipp8u* pDst, int dstStep, IppiSize roiSize, IppHintAlgorithm hint ))
This function scales pixel values of the source image ROI and converts them to the destination data type according to the following formula:

dst = saturate_to_dstType(src * mVal + aVal)

regards, Igor

Hi Igor,

BMart1 — Fri, 10 Aug 2018 10:58:13 GMT

Hi Igor,

Thanks for the suggestion. ippiScaleC_16u8u_C1R does work. The results are not exactly the same because it uses floats, which probably also makes it slower. I ended up writing the functions with intrinsics.

Bruno