[XviD-devel] minor changes / todo list

Michael Militzer michael at xvid.org
Wed Jun 25 14:51:56 CEST 2003


Hi,

Quoting "Hans K. Rosbach" <hk at circlestorm.org>:

> Fist: Thanks for your comments, they cleared up a few things.
> 
> > It could be helpful if we add a threshold to your idea and only reencode
> > a frame if we horribly missed the desired frame size. I just wonder how
> > frequently this happens (hopefully just very seldom). In this case the
> slow
> > down wouldn't be severe but also the benefit might be questionable...
> 
> allowedoverflow / allowedunderflow was that threshold. So my suggestion
> was only meant to correct those frames that are way off the desired size.

oh, sorry. I wasn't looking close enough at your code.
 
> We would not get all the benefits of 3-pass, but I do think that this might
> increase overall quality a little. The question is just: Is it enough?

yes, overflow could be reduced. But how much will it influence quality?
Usually overflow is no problem, because the rate control isn't compensating
overflow/underflow that occurred immediately but distributes the error over
a longer sequence. So often, over- /underflow is self-compensating over a
couple of frames, or can be compensated by just a small change in quantizer
for a certain period of frames.

So your idea should only be really helpful, when the current quantizer 
estimation is completely wrong, when heavy over- or underflow occurs that
would lead to big quantizer changes. Hopefully this does not happen with
our current rate control algorithm, but maybe we can find a clip where it
could...

> I imagine this functionality could also be extended to support a minimum
> wanted psnr value for the frames (with a threshold of how many re-tries
> the codec should attemt to increase psnr). This would also need more
> intelligence of course.
> 
> The "minimum"/fixed PSNR is something I have been wanting to test for some
> time now, but so far I know only Mpeg-2 encoders that do this.

hm, min. PSNR says not much. PSNR is no good metric for perceived visual
quality, so a guaranteed minimum PSNR does not mean a lower-bound for
subjective quality. However it could be helpful to prevent really ugly
looking frames, but I fear it could become impossible to guarantee a
minimum PSNR when aiming for VBV compliant rate control. Well, it will
depend on the buffer size...

bye,
Michael


More information about the XviD-devel mailing list