[vorbis-dev] Tarkin video codec?

Gregory Maxwell greg at linuxpower.cx
Tue Dec 19 16:16:34 PST 2000

On Tue, Dec 19, 2000 at 03:50:31PM -0800, Ralph Giles wrote:
> done some test work[1] but never released the code. I've been urging
> people to just start from scratch; there's enough info in the list
> archives to duplicate the part they've written and a number of new

Perhaps someone should assemble a list of unpresued directions?

My most recent thought:

Using some sort of 'space filling' curve to reduce the 3d video data to
lower dimensionality before or after 2d transform to overcome any pitfalls
of using a 3d transform directly (the sharp discontunity between the
resoultiof of {x,y} and {z} axises gave me problems when expirmenting with a
3d transform directly).

> Beyond that, I'd suggest studying the design of the vorbis codec, in
> particular the codebook implementation; I understand it's smart. :) Then
> experiment with different partition and transform schemes to find

A thought here, Monty's code is, as usual, increadibly flexible and well

Correct me if I'm wrong, but one advantage of a 'huffman coupled vector'
coder is that it can perform simmlarly to an arithmetic coder. (No? Main
advantage of of an arithmetic coder is that it works more optimilly when the
symbol probabilities are not negitive powers of 2, a huff coded vector coder
can represent such sets using negitive power of two probability symbols with
non-negitive power of two 'subsymbols')

--- >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.

More information about the Vorbis-dev mailing list