2013-11-15, 23:23
Looking at xbmc code, CalculateYUVMatrix() in xbmc/cores/VideoRenderers/VideoShaders/YUV2RGBShader.cpp makes me very suspicious. I think the translation by -16/255 should not be done and the order of scaling to full range and translating Cb and Cr should be reversed.
Overall, I don't like the way it's structured. Clearly the devs prefer full range output, but I think limited range is better quality. Full range with dithering or higher bpp may provide some benefit when processing is applied, but neither is currently supported.
With the limited range output messed up, it is difficult to demonstrate why it is better than full range. Good news is that if this is the problem, it should be easy to fix.
Overall, I don't like the way it's structured. Clearly the devs prefer full range output, but I think limited range is better quality. Full range with dithering or higher bpp may provide some benefit when processing is applied, but neither is currently supported.
With the limited range output messed up, it is difficult to demonstrate why it is better than full range. Good news is that if this is the problem, it should be easy to fix.
