And that's all for lossless compression. For lossy compression, there are lots of unexplored areas.

I still like the FrameFree approach to video compression. Motion estimation decomposes sequences of images into a set of layers which move and morph, much like the way 2D movies are converted to pseudo-3D. This is hard, but can be done, and there are several software packages for it, of varying quality. Decompression is done by morphing each layer between keyframes, warping a texture-mapped mesh in a GPU. Decompression can be done to any frame rate, including ones much higher than the original. This approach is often used to generate ultra-slow motion.

(FrameFree was demonstrated around 2006, but the company behind it, Kerner Optical, went bust for unrelated reasons.)