[XviD-devel] 2pass curve compression

Edouard Gomez xvid-devel@xvid.org
Fri, 9 Aug 2002 19:36:18 +0200


Dirk Knop (dknop@gwdg.de) wrote:
> Hi,
> 
> yetserday I talked some time with Foxer about "his" alt CC curve 
> compression algorithms.
> In th end, the descriptions can be made a little more understandable in 
> a less-technically language:

The   code,   even   if    it   lacks   some   comments,   is   quitly
understandable. When  i ported the  code for the independant  vbr lib,
the code  revealed me what it  does without too  much difficulties. On
the other hand some parameters were obscure... but anyway, i'll try to
make a good document  on how the vbr code does the  job (could be used
for both vfw and whatever project based on the vbr lib)

> 
> I would like to see new default values implemented into XviD [...]
> 

I will apply  your defaults in transcode too, so  every XviD user will
have the same result for the same movie and same set of paremeters :-)
thx

> I'd like to have a linear scaled curve with a high and lowpass, some 
> overflow treatment which works like it now works (scene based), and some 
> mechanism to prevent too small frames to get quantized more ;) Maybe a 
> reaction delay for the oveflow treatment as well.

You could also try to  build an heuristic function f(mblocks, ublocks,
iblocks)  that  would  return  a  hint on  how  the  quantizer  should
evolve. The  above function should use  a sliding window  to have both
past (overflow,  need to  be controled) and  future estimation  (so we
would know when  a scene becomes static or with  action). This is just
an idea I had when looking at the code during the lib port. But I know
this heuristic  function could help a  lot but is  rather difficult to
tune.


> Ok, just wanted to share thse results.

Thx

-- 
Edouard Gomez