[Ffmpeg-devel] VP6 encoder

Michael Niedermayer michaelni
Mon Oct 23 21:36:55 CEST 2006


On Mon, Oct 23, 2006 at 04:55:36PM +0200, Derk-Jan Hartman wrote:
> On 23-okt-2006, at 14:20, Stefan de Konink wrote:
> >On Mon, 23 Oct 2006, Eduardo wrote:
> >>At 05:35 23/10/2006, you wrote:
> >>>>
> >>>>Snow and dirac maybe? (help wanted to get it finalized)
> >>>
> >>>Horrid performance, even more patent claims, and compression  
> >>>presently
> >>>much worse than h264...
> >
> >Dirac has hardware, and no patent claims. That is claimed by BBC.
> There are actually several patents that are claimed BY the BBC, so  
> that the BBC can defend the project against other companies sueing  
> based on patent claims. In comparison to the other codecs the wavelet  
> field holds much less claims overall however.
> Also I saw the new versions and the Dirac hardware at IBC and was  
> REALLY impressed. There has been made huge progress and visually it's  
> almost on par with H.264

it would be interresting to see some comparission of x264 (8x8dct+cabac+all
the other goodies), snow (with iterative me), mpeg4 asp (with all the things
like trellis quant enabled) and dirac in PSNR per bitrate and subjective 
quality per bitrate (double blind testing style)

ive never tested snow against h.264, so i cant speak about that but when
compared against mpeg4 asp, snow had noticable better quality

i tend to have a weird feeling if i hear people claiming "almost on par"
and such, at least vc-1 and Chinese AVS have their own "as good as h.264"
claim ...

> The problem is still the speed of course. It's one of the major  
> reasons they are also working on hardware I guess.
> I would like to point out the true commitment the BBC is making here.  
> Many people don't seem to realize that the BBC has vowed to make  
> their entire production chain Open Source wherever possible. If they  
> have to by developing everything themselves. MXF/GXF/Dirac,  

why exactly dont they use nut or matroska, or hell even avi, mpeg-ps or mov
MXF/GXF are not really containers i would use if i had the choice


> >>P.S. Is there any paper describing snow codec? Or is everything on  
> >>source code?

source code sorry, if someone wants to help documenting the algorithms
or wants to try other algorithms, all that is very welcome ...

Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

In the past you could go to a library and read, borrow or copy any book
Today you'd get arrested for mere telling someone where the library is

More information about the ffmpeg-devel mailing list