[XviD-devel] VHQ
Christoph Lampert
chl at math.uni-bonn.de
Wed Feb 12 15:33:11 CET 2003
On Thu, 13 Feb 2003, Radek Czyz wrote:
> I think I like the way it is now - if cbp = 0, we count on SAD. If
> not, we count bits.
hm, but where is there difference between minimizing vector bits with
fixed dct-bits between cbp!=0 and cbp==0 ?
In the second case, there is no chance to lower number of dct-bits, okay,
but still if we have several possibilites with same number of DCT-bits,
vector bits don't give a clue which to choose, and minimizing them would
be bad.
I guess your method works, because due to the nature of video signal,
it's less likely that DCT bits stay constants and vector bits done, unless
everything is 0 from the beginning already, but still I have the
impression that a method
* "Optimize DCT bits under the condition of total bits not getting larger
than what ME had found",
instead of
* "optimize total number of bits"
would be worth looking at.
I'll check a few samples how "# of bits" evolves during this search
and see if it would make difference.
gruel
More information about the XviD-devel
mailing list