[XviD-devel] minor changes / todo list

[Michael Militzer]

Hi,

Quoting skal <skal at planet-d.net>:

> > Even if you want very low overhead there are smarter ways to adapt
> > the deadzone I think (but Id have to think a lot harder to come up
> > with one :).
> > 
> 	the simplest i can think of is using retroactive feedback
> 	(statistics) for mapping the effect of "widening the dead-zone
> 	that much saved me that amount of bits and trashed the PSNR by
> 	that amount". It's usually the solution when one don't have a
> 	better solution. "Let the system adapt!". :)

ok, so the main benefit from varying the quantization bias should be the
increased dead-zone resulting in more coefficients quantized to zeros,
right? So yes, more zero coefficients can have a positive effect on the
bits needed to code a block but is it really worth the decrease in PSNR?
Or better: how to determine when it's worth it and when not? And that's
a R-D problem again which when solved properly introduces significant
complexity.

I'm not sure whether your proposal of "Let the system adapt!" will be
really that successful: The R-D characteristics of 8x8 blocks within one
frame are different, so a bias that was good for the previous block does
not have to be good for the current block. Ok, maybe there is a certain
spatial correlation in R-D characteristics of neighbouring blocks which
may help, but I wouldn't rely on it. I really doubt that the result's
could rival a simplified Trellis approach...

bye,
Michael