Re: Re: Encoder-decoder interface question.
"ChristianHJW" <[email protected]> Fri, 18 Oct 2002 12:00:56 +0200
| Newsgroups | gmane.comp.multimedia.ogg.tarkin.devel,gmane.comp.video.uci.devel |
|---|---|
| Message-ID | <[email protected]> |
"Kondoros Attila" <[email protected]> schrieb im Newsbeitrag news:[email protected]... > I have a small question: 3D wavelets are out of the > question, don't they? I thought that we pursue 2D > wavelet transform and a motion compensation scheme for > now. Am I wrong here? But if 3D wavelets remains then > the interface must be able to easily pass trought > multiple frames. I don't know if other codecs are > doing this right now... > As long as the "frame-in -> stream out" methodology > remains, the interface should stay the same even if > those internal algorithms will improve. > Attila Here we are, exactly this was the background of my question about UCI <-> Tarkin compatibility. The current VfW API allows only one frame in - one frame out . This is already big problem for most modern MPEG4 codecs using b-frames, because they have to load more than 1 frame into cache to be able to make the decision whether to use a b-frame or not, and of course also to calculate it from previous and next frame then. Only chance to support that with VfW and AVI was to use a 'hack' / workaround where the codec loads a few frames into cache at the very start of encoding and marks those frames as 'bad' or 'dropped' . UCI allows more than 1 frame to be loaded before outputting them, but i was wondering if this was suffcient for a 3D wavelet codec ? Best regards Christian Sites : http://mcf.sourceforge.net http://sf.net/projects/mcf http://uci.sourceforge.net MCF mailing lists : news://news.gmane.org gmane.comp.video.mcf.general gmane.comp.video.mcf.devel gmane.comp.video.mcf.mplayer gmane.comp.video.mcf.announce gmane.comp.video.mcf.mpc gmane.comp.video.uci.devel Soon : www.corecodec.com --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to '[email protected]' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.