[Vorbis-dev] improving vorbis compression with arithmetic coding and by exploiting temporal redundancies

[Balatoni Denes]

Hi!

csütörtök 01 szeptember 2005 23.06-kor Dimitry Andric ezeket a bolcs 
gondolatokat fogalmazta meg:
> On 2005-09-01 at 23:07:18 Yann Guidon wrote:
> > If it was known in advance, then the efforts were fruitless.

For me it was not known in advance - and probably the exact degree of the 
suitability of mdct for said purpose was not widely studied. AAC too has 
temporal prediction in the standard (and it is AFAIK much simpler).

> > Implementing arithcodes is an easy and effective method.
>
> However, isn't arithmetic coding encumbered with numerous patents?
> This might have been the reason that it was never used in Ogg
> Vorbis...

Yes it is said to be patent encumbered, but the reason why huffman and vq was 
used is in my humble opinion computational complexity. Arithmetic coding 
could be substituted with range coding (very similar, but old enough not to 
be patent encumbered), but it is almost as computationally intensive afaik. 
And as it seems vq and huffman performs well.

> There are probably lots more interesting and useful compression
> improvents that could be added, if only they weren't patented by the
> large media corporations (e.g. Philips, Thomson, etc.)

Yes, just read the AAC standard.

bye
Denes

-- 
---
What kills me, doesn't make me stronger.