[XviD-devel] Idea for encoding iframes (or images for that matter)

[Jan Lucas]

Christoph Lampert wrote:
>>>in this paper <http://citeseer.ist.psu.edu/dequeiroz00fast.html> could
>>>be interresting. (The FLT is a modified DCT transform, with almost
>>>wavelet like behavior, but without the performance hit)
> I don't want to spoil you enthusiasm, but a Citeseer paper from 2000 that 
> has received only 2 weak citations since then isn't a very promising 
> source for a revolutionary new video codec. 

Who said revolutionary? I just said, it could be interessting to try.
QPEL and GMC aren't helping that much either on most stuff, while still
having a noticable performance hit. FLT could be a similar small
improvement. And if we just find it isn't helping at all, we just
learned why it got only the 2 citations.

>>So I suggest making an experimental branch, which we might call...
>>er.... ermm.... :( I don't know, "project junio"? (reference to proj.
>>mayo ^^;; ) Anyway. I think that we should start a mailing list,

I don't think so. Starting a mailing list for something that is just a
small idea doesn't really make sense. Get a XVID source copy, add stuff
you wanted to try to the code, and if you really get better performance
than xvid, THEN you can start a mailing list. Otherwise you will just
get something similiar to ogg tarkin, where a lot of discussion happend,
 but afaik no working, useful code was written.

>>webpage dunno, for XVid developers wanting to improve the algorithm. A
>>warning: We should make sure the algorithms we use are NOT under patent
>>research. I've seen MANY of those in citeseer.
> 
> 
> Then you should forget XviD. MPEG-4 is covered by tons of patents, and you
> can't get rid of them by stripping it down, it's just too much. Others
> have tried to start a MPEG-like but patent/royalty-free codec, e.g. the 
> whole AVC/H.26L team, but in the end it was again patents everywhere. 

Yeah, if you want to go patent-free, using XVID as a starting point is
stupid. But if you are just searching for a fast playground to try out
some stuff, without caring about patents, XVID might not be such a bad
starting point.
(Theora or Dirac are nice too sure)

							Jan