@aexiruch as is it isnt really useful for compression. That said the idea in general is used in already well established compression approaches. This is nothing novel in the general case.

I remember we were doing this sort of "cheap" compression (if you can even call it that) back in the early 90s. You basically just assign less resolution to different parts of the color encoding for example saturation vs luminescence.

@freemo Yeah, I just wondered if it was technically feasible / an interesting hack. I understand the majority of the more old-school compression algorithms, but I never came across one that so blatantly exploited the way the human brain processes single elements of color in a luminance field... Most reproduce something even a computer will trivially recognize as a similar image, this one goes beyond that.