Can you check new the IPP 7.1 interfaces on the resize function, and see how it works for you?
For scale the image pixel, I think you may call a separated function after the resize.
You are right about ippiScale -- it is for converting individual pixel values to different bit depths. For example, ippiScale_8u32f_C1R with vmin of 0.0 and vmax of 1.0wouldconvert the 0-255(integer) data from the single channel image to floating point values between 0 and 1.
There are many choices for interpolation mode since there are tradeoffs between quality and speed. Since you specifically don't want interpolation, I wonder if nearest neighbor interpolation mode might be closer to what you want. (IPPI_INTER_NN for ippiResizeSqrPixel.) The idea behind this algorithm sounds very close to your description, and at least in my tests I'm not seeing interpolated values using this mode.
This is a Feature Request and please consider two additional modes:
IPPI_NOINTER_UPSAMPLE ( scale factors: 2x, 4x, etc )
IPPI_NOINTER_DOWNSAMPLE ( scale factors: 0.5x, 0.25x, etc )