project extension proposal

Hi all.
Here, at TILAB, we started implementation of an optimized H264 Codec.
I have joined the list just to feel what is happening around, but unlikely I
will be able to help much (I wish I could :).
Our decoder is now 10 times as fast as the reference one, with a very simple
scenario (I,P,Cabac and deblocking). (no bullshit)
And it is really 1000 times simpler!
In 14496/2 we chose to go from scratch for encoder, and keep M$ as a
starting point for decoder.
In 14496/10 this looks almost impossible, as the code is really unreadable,
which very often is a good companion of inefficiency.
Very likely, in a fast implementation, you will need border-padded frames
for
a fast Motion Compensation...
And anyway, at last testing is much easier, as there is a nice, nice, nice
and well
specified transform (Hadamard) and decoder is bit exact.
So my suggestion is go for a brand new implementation.
It does not require much more than one (skilled) man month work and the
results
are fairly impressive.
Good luck.



-------------------------------------------------------
This SF.net email is sponsored by:
The Definitive IT and Networking Event. Be There!
NetWorld+Interop Las Vegas 2003 -- Register today!
http://ads.sourceforge.net/cgi-bin/redirect.pl?keyn0001en