It should be possible to implement dithering at the selector finding step by adding an ordered dither pattern to the data.
Possible issues:
- PSNR favors non-dithered data, so error value calculation may need to be changed, and benchmarking may be difficult
- blocks with endpoints of different hues may end up looking strange with dithering applied
It should be possible to implement dithering at the selector finding step by adding an ordered dither pattern to the data.
Possible issues: