NPP JPEG Compression problem nppiDCTQuantFwd8x8LS_JPEG_8u16s_C1R parameters

Hi,

I am currently developping a GPU JPEG compression system, but I have a problem using the nppiDCTQuantFwd8x8LS_JPEG_8u16s_C1R function.
For example I have a 16*16 greyscale image, which parameters srcStep, dstStep and ROISize have I to specify for the forward & inverse DCT Functions ??

Tahnk you.

Here is some proof of concept code I posted in another entry. Note the comments about the broken quantization functions in NPP 3.2. I haven’t tested the correctness of these functions in CUDA 4.0.