Sticking with image compression, see Quite Okay Images. It treats each pixel as three numbers and expects mostly small changes. Recent pixels get hashed and can be referenced in a few bits. This is enough to compete with PNG filesizes, an order of magnitude faster, while handling each pixel exactly once.

[–] MinekPo1@lemmygrad.ml 2 points 11 months ago

though note than lossy formats , like JPEG which was used here , do use Fourier transforms , which are very intense trigonometry . IIRC PNG doesn't use trigonometry either , ~~though I'm not entirely sure~~ yup PNG uses DEFLATE after some filtering , so no sine there I believe