> rather than doing basic classic transforms like converting to YUV420 What will converting to YUV420 achieve though, except for 4:2:0 chroma subsampling? YUV

> rather than doing basic classic transforms like converting to YUV420 What will converting to YUV420 achieve though, except for 4:2:0 chroma subsampling? YUV has little basis in human perception to begin with, it's a color television legacy model used for compression. There are much better models if you want to extract the perceptual information from the picture.