Balatoni Denes
2005-Aug-31 11:03 UTC
[Vorbis-dev] improving vorbis compression with arithmetic coding and by exploiting temporal redundancies
Hi! If you were looking foe a paper about improving vorbis compression by using arithmetic coding and exploiting temporal redundancies (i.e. repeating sound) you might have a look here: http://web.interware.hu/rudas Three ogg vorbis files were compressed losslessly, with compression ratios between 2 - 8 % and compression time about 1/500 realtime (i.e. slow). Theoretical upper bound of the compression for real music is estimated to be between 10 and 20 % - further improvements to the methods are possible. bye Denes -- --- What kills me, doesn't make me stronger.
Yann Guidon
2005-Sep-01 13:51 UTC
[Vorbis-dev] improving vorbis compression with arithmetic coding and by exploiting temporal redundancies
Hi ! Balatoni Denes wrote:>Hi! > >If you were looking foe a paper about improving vorbis compression by using >arithmetic coding and exploiting temporal redundancies (i.e. repeating sound) >you might have a look here: > >http://web.interware.hu/rudas > >Three ogg vorbis files were compressed losslessly, with compression ratios >between 2 - 8 % and compression time about 1/500 realtime (i.e. slow). >Theoretical upper bound of the compression for real music is estimated to be >between 10 and 20 % - further improvements to the methods are possible. > >interesting read, indeed ! But this method is not going far, and the author explains why : " However, the most striking observation is that the MDCT transform, as used currently, is not well suited for finding and exploting temporal redundancy. The most likely cause of this is the MDCT's lack of translational invariance." If it was known in advance, then the efforts were fruitless. Implementing arithcodes is an easy and effective method. But prediction by looking at the past is already difficult, but the blocking due to the MDCT behaviour makes it more difficult. Maybe wavelets would have been more practical. I also wonder how a decoder would behave in the presence of errors, and how long the generated "artefacts" would last. Errors (if the Ogg container's CRC is not verified) would have effects up to 25s, according to the text. But it's interesting anyway :-) Please come back with other experiments !>bye >Denes >YG