[vorbis-dev] video codec

[David A. Gatwood]

On Sat, 10 Feb 2001, Holger Waechtler wrote:

> > Thats not the point, its the user's responsibility to present the image as he
> > would like them compressed. Suggesting using a noise polluted source to test is
> > doing things backwards... if it removes them its ok and if it doesnt remove them
> > its still ok, well that was a usefull test. Test with an image sequency which
> > naturally has high frequency components.
> 
> Noisy high frequency components are a real problem (and will be for any
> comprssion codec); they generate lots of uncorrelated coefficients, which
> are hard to compress. Perhaps we can get out of this by using a 'texture
> syntesizer': analyze some typical frequency characterisics of the
> (untransmitted) noise and resynthesize this by adding shaped random noise
> to the coefficients on the receiver side. The simplest way would be to
> save a single noiselevel-per-scale value.

Here's something I'd like to see in a video codec.  When I encode video
with e.g. MPEG or RealVideo, there's one thing that's really obnoxious. 
If I have a still portion of an image and a moving portion (for example, a
title over moving video), basically every low bitrate codec I've seen
sucks at representing it.  The title sort of blurs into the video,
resulting in unreadable test and a rather poor representation of the
background video as a consequence

I'd like to see a codec look at multiple frames and look for common
components, such as those titles, and if it sees parts that change almost
not at all across two or three frames, store those parts as a super frame,
average them out of every frame before encoding (smear the surrounding
areas in) then re-superimpose the still portion on playback.  That would
both make the titles more readable and eliminate that white smear that
makes the background footage look pretty bad.

Any thoughts on that?  Obviously, it'd be more CPU intensive than normal
encoding, so it would probably need to be optional, but if done correctly, it
could really improve a large number of video clips.

Thoughts?
David

---------------------------------------------------------------------
                    Check out my weekly web comic:
                     http://www.techmagazine.org

--- >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-dev-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.