Hi demo pages http://web.mit.edu/xiphmont/Public/theora/demo.html http://web.mit.edu/xiphmont/Public/theora/demo2.html discussed opportunity to add range coding there is already some xiph code: http://svn.xiph.org/trunk/ghost/libentcode/ and Timothy B. Terriberry wrote some time ago ==I'd estimate that you'd save ~12% by switching from Huffman to something arithmetic coding-based for video (based on the experiments with H.264... Theora has more flexible Huffman coding for some pieces, and less flexible coding for others, so it probably averages out on the balance). == now as tuning is finished, will anyone attempt to perform tests ( as it follows from "based on the experiments with H.264" it should not be such difficult to add changes ) how much theora encoder will gain from range arithmetic coder on real clips ( using for example http://media.xiph.org/ )? Or how difficult do you estimate to add range coder? Regards Sergey
Hi btw in addition to range coder http://svn.xiph.org/trunk/ghost/libentcode/ there is schroarith.c in dirac Schr?dinger codec which has arithmetic coder implementation. I'm not sure, but still seems that it is relatively easy ( matter of days, maybe weeks ) to replace current encoder/decoder with minimal entropy coding with above mentioned code. after there is an implementation ( not having time to try by myself now though, but I will try someday if no one will attempt ( even if I miss something ( like additional effort for range coder ) at least will learn how it goes ) it would be possible to compare relative efficiency on real clips. at least I hope that for custom projects ( in case on no 'quick' updates to spec ) there will be already free codec which beats or ties with h264 :) as it looks like further steps (probable addition of mdct etc ) might take year or even more to ratify specs and to implement. Regards Sergey