why fractal video codecs did not become mainstream?

[ckmox]

i read somewhere that the holy grail of video compression is fractal video compression its main advantage is that its resolution independent just like vector images, so no need for resolution wars like 720p vs 1080p, and the resulting file sizes are very small too, we got fast computers now to encode fractal videos so anyone can explain what is/are the reason/s fractal video codecs are not being considered by experts now?

[foxyshadis]

Were you reading something written in the 90's? It was a good idea for its time, and it led to some interesting advances, but not only was it obnoxiously slow, it ended up hitting a wall and losing significantly to the state of the art that it competed with (H.263 and MPEG-4 SP). Most of the big splashy examples of its "holy grail" efficiency were because of fitting specific fractals to test images, which can never scale to the real world. The efficiency of fractal coding only exists for intra anyway, which is only a small part of video coding; it's worse than DCT for residuals, and progressively better vector-prediction in MPEG-4 codecs had a much larger impact. With H.264 and HEVC, much better intra encoding and intra prediction has led to them being far better than the best fractal coded pictures, and so it's now only a footnote in history.

[ckmox]

ye i was reading some 90's articles about fractal compression, thanks for the quick answer!

[benwaggoner]

Were you reading something written in the 90's? It was a good idea for its time, and it led to some interesting advances, but not only was it obnoxiously slow, it ended up hitting a wall and losing significantly to the state of the art that it competed with (H.263 and MPEG-4 SP). Most of the big splashy examples of its "holy grail" efficiency were because of fitting specific fractals to test images, which can never scale to the real world. The efficiency of fractal coding only exists for intra anyway, which is only a small part of video coding; it's worse than DCT for residuals, and progressively better vector-prediction in MPEG-4 codecs had a much larger impact. With H.264 and HEVC, much better intra encoding and intra prediction has led to them being far better than the best fractal coded pictures, and so it's now only a footnote in history.
Moreover, I don't think there were ever any truly "resolution independent" fractal encoding implementations. They used something called (IIRC) the Iterated Transform which was fractal inspired, but wasn't purely fractal.

I worked on this at a partner company of Iterated Systems back in the late 90's. We never got it to be consistently better than even J2K, and the encoding time to get to efficient compression was CRAZY.

HEVC IDR frames would blow it out of the water.