They use wavelet transforms for jpeg 2000. Natural images tend to be sparse with respect to wavelet transforms.

I think JPEG is too. In school, we implemented something like JPEG compression using some Fourier transform stuff that reduced the bitmap size by 90%+ and was still fairly decent looking. It was an image handpicked for suitability under compression, I'm sure. A sunflower, I think?

I had a similar kind of lab in college. I ended up testing it on an image hand-picked for performing very poorly under JPEG compression, a GBA Pokemon game screenshot. Chroma subsampling alone visibly degraded the sharpness of brightly colored objects. Strangely chroma subsampling degrades photographic imagery far less, and helps less sophisticated algorithms (like JPEG) allocate limited bits better.

Ah, you were too young to use the correct image for all testing purposes: Lena Forsén.

The image processing community has moved away from Lena. It's a sexist nude crop from Playboy Magazine. It's uncomfortable and gross to use that in a professional setting where we're meant to be welcoming and inclusive.

https://pursuit.unimelb.edu.au/articles/it-s-time-to-retire-...

You're right, and I'd never considered it before. Thanks for the link.