Coding 2025-03-30
All right, I got some testing done in a notebook, and figured out what was wrong with my Bayer dithering implementation; now that I've fixed it, it can be excellent for some images.
Riemersma dithering was fun to implement, but so far I haven't found an image where I feel that it really shines. There's some where it does the best job out of what I have easy access to so far, but I'm guessing that could change once I implement Atkinson dithering.
Pillow, as I might have mentioned before, provides Floyd-Steinberg, and sometimes this is fine, and sometimes it kind of isn't. Hence why I'm interested in other forms of error diffusion.
Anyway, I ought to wind down now. Hopefully tomorrow, I'll have time to wrangle matrix-based error diffusion out of Numpy's API.
Good night.