[Matroska-devel] CodecPrivate for MPEG-1 and MPEG-2 video

Moritz Bunkus moritz at bunkus.org
Tue Nov 16 18:07:37 CET 2004


> IIRC thats not correct, because sequence header can change many times 
> during the track/movie. If you consider the right solution being the one 
> thats ideal for concatenation or editing, then the sequence header 
> should be left inside the frame IMO, even if this means a bit less 
> saving on overhead compared to MPEG container.

I don't think that the sequence header has such a big impact on
overhead. And yes, they may occur more than once. If we decide to throw
them all out then we have to start using CodecState, too, for the places
where the sequence header changes. I don't even know the exact layout /
size of a sequence header. That's why I'd like some input from spyder

> But, i am pretty sure all this has been discussed many times already, 
> even here on this list.

No, it hasn't :) That's why I've brought it up. (Nearly) the same
discussion will follow one day for MPEG-4. The thing is that MPEG-2 and
MPEG-4 seem to be stored differently in MP4, though, so we can't just
say "let's do it like the MP4 guys" either. MPEG-2 in MP4 simply
contains the sequence headers in the frames, and there's no "private
data" in the ESDS (elementary stream descriptor).

For MPEG-4 it's the other way round. The ESDS contains private data
which is missing in the frames. This is a big difference compared to how
it is stored in AVI: In AVI, each MPEG-4 I frame contains this private
data (if I'm not mistaken).


If Darl McBride was in charge, he'd probably make marriage
unconstitutional too, since clearly it de-emphasizes the commercial
nature of normal human interaction, and probably is a major impediment
to the commercial growth of prostitution. - Linus Torvalds

More information about the Matroska-devel mailing list