[MPlayer-dev-eng] [PATCH] synch to new x264 revision

Guillaume POIRIER guillaume.poirier at ifsic.univ-rennes1.fr
Fri Sep 24 10:52:40 CEST 2004


Le ven 24/09/2004 à 03:14, Jeff Clagg a écrit :
> I finally did a simple test of this, and obviously frameref can be a
> big win on flashing scenes. I made a simple test sequence of alternating
> frames. The ringing artifacts are due to me saving the images as jpegs
> and then quickly deciding it's not important enough to go back and do
> again. I encoded 27-frame videos, of these two images alternating back and
> forth on even and odd frames, using frameref=[1-8]. Summary of results:
> 
> frameref  size (bytes)
> --------  ------------
> 1         833948
> 2          73608
> 3          74928
> 4          74928
> 5          74928
> 6          74928
> 7          74928
> 8          74948
> 
> As expected, starting at frameref=2, the first two frames are iframes
> (coded sizes 31256 and 30214 bytes) followed by 25 p-frames.

Good to know frameref is working as expected...
May I suggested the same test on a movie-real-world scene? How about a
traveling with trees in the foreground that hide the background?
... or maybe it's already taken care of by regular MPEG-4?
I'm not the movie expert here.

I did make some "explorations" with Romain Dolbeau's mencgen
(http://www.caps-entreprise.com/private/dolbeau/lavcGenetic/lavcGenetic.html).
(Nice tool, you should try it ;-)

The test is 3000 frames long and is taken from a clean recent cartoon.

fullinter improves PSNR 40% of the time

deblockalpha doesn't help 73% of the time, 
helps when set to 1 13% of the times

iframe interval does need to be increased. If helps when set higher than
100 50% of time

qcomp helps when set to 0.6 50% of times

subq helps when set to 5 or 3 50% of times
and helps when set to 0 33% of times

qblur helps when set to 0.5 53% of times and when set to 0.3 27% of
times

iframes help when set to 2 47% of times, when set to 3 33% of times

Deblockbeta doesn't help 73% of times

deblock helps 50% of times


So don't take my word for it, test it, as this tool bases its decisions
on PSNR.

Regards,
Guillaume




More information about the MPlayer-dev-eng mailing list