[vorbis] bitrate peeling

Christian.Buchner Christian.Buchner1 at epost.de
Sat Nov 9 23:00:39 PST 2002



> I really never did get an answer to my original question (Above). I have to 
> assume this must mean that the peeling support in ogg is really just smoke 
> and mirrors. 

> A feature as important as this should have had tools to do it by now. Perhaps 
> something is broke in the spec and nobody yet wants to admit it.

Well I believe the encoder will have to be improved in such a way that it tells
the peeler how to remove bytes such that quality will be affected minimally.

If you start ripping away data indiscriminately from audio frames (like removing
50% from each ogg frame), then you will get quite a poor quality in some parts
of the song. However if the encoder "hinted" the ogg file with a quality vs. bitrate
metric, the quality loss would be less severe. Ideally the quality loss caused
by peeling would be distributed evenly over the whole duration of the song.

I am sure the spec does not define any such quality metric and how it would be
transmitted in the stream (in a separate multiplexed ogg stream maybe?).

And there's another issue. If you peel more in some parts and less in other parts,
you need some kind of bitrate reservoir, especially if you try to peel to a constant
bitrate. I do not know about such a facility in Vorbis currently or how this would
have to be implemented.

Another thought: Maybe it is possible to develop a peeler which has its own
psychoacoustic model so it can intelligently decide where to rip off bits and where
better keep 'em. So, concluding: Writing a good peeler could be almost as complex
as writing a good encoder!

Most of the issues mentioned have not yet been addressed. And not enough people
REALLY know the internals of Ogg  to do this work (I confess, me neither!). So it
will have to be done by the hardcore Xiphoporus people.

There has been lots of scientific research done about MP3, ACC and other industry 
standards. But I've never seen papers been published about Ogg so far - also no
good introductory articles summarizing the main concepts ;( And that again leads to
less people learning about the internals Vorbis. I wonder what could be done to
improve this situation.

Christian

<p><p>--- >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 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis mailing list