What is the Sate of Snow?

[iwod]

Since its Introduction ( at least on Doom9 ) in 2004, very little has been said on it.

Has it been abandon?

Where are we going with it? Or Dirac and Snow are too similar?

[Dark Shikari]

Despite its lack of B-frames and a near-complete lack of development over the past half-decade, it still seems to be competitive with Dirac, probably due to its better entropy coding and iterative motion search. Of course that isn't saying much (http://saintdevelopment.com/media/)...

It still has all the practical problems of OBMC wavelet formats: too blurry, too much aliasing, crappy intra coding, very high decoding complexity, difficult to implement variable motion block sizes. Dirac has solved precisely zero of these.

A lot of the problem is theoretical: while one can always approach a format with trial-and-error to look for ways to improve coding efficiency, one still needs a basic theoretical idea behind why such a change would (or would not) improve things... and most of the core problems with wavelets 5 years ago are still core problems with wavelets today. And yet a lot of the wavelet research is going into producing ever-more-PSNR-optimal solutions, to the point that they're basically just overfitting transforms for PSNR at the cost of everything else.