[FFmpeg-devel] [PATCH] Issue 543, clarification of -bt in docs

Michael Niedermayer michaelni
Sat Jul 26 04:18:00 CEST 2008


On Sat, Jul 26, 2008 at 04:02:33AM +0200, Michael Niedermayer wrote:
> On Fri, Jul 25, 2008 at 03:21:44PM -0600, Loren Merritt wrote:
> > On Fri, 25 Jul 2008, Jason Garrett-Glaser wrote:
> > 
> > >> btw, one thing i wanted to ask you since a while already ...
> > >> Are you interrested in merging the x264 & lavc ratecontrol code?
> > >> The reason why iam asking is that IMHO the current state of it being
> > >> "forked" in x264 and being improved there isnt ideal. And that will
> > >> become just worse over time as people improve both independantly ...
> > >
> > > The main problem I see is that x264's ratecontrol code depends heavily
> > > on an Xvid-style low-resolution lookahead, especially the VBV code.
> > > I'm not familiar enough with LAVC's ratecontrol to know whether LAVC
> > > has similar, but if it doesn't, that is a dealbreaker.
> > 
> > lavc instead runs fullpel motion estimation before ratecontrol. The 
> > fullpel sads should be usable for the same purposes. I don't know which 
> > is more accurate at predicting complexity.
> 
> This isnt correct, we run the complete motion estimation before 1pass
> ratecontrol. Prior to the main motion estimation there is an optional
> pre motion estimation pass that does pretty much what normal ME does but
> bottom to top.
> 
> Also both ME passes have fully user configureable comparission functions,
> so SAD,SSE,SATD, ... are all there if the RC code has any special preferance.

Note, though, looking at the code now it seems we only export the fullpel SSE
scores, but it would be trivial to export the final subpel scores if needed.

I think the reason why it isnt done currently is that i wanted to add a
new user defineable comparission function for what the ratecontrol code
gets, but i never did implement that. And just passing the output
from whatever the user selected as subpel ME comparission function to the
RC code likely would cause problems. Still the subpel score as they are
could trivially be used as long as the comparission function is what the
RC code expects ...

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I count him braver who overcomes his desires than him who conquers his
enemies for the hardest victory is over self. -- Aristotle
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080726/b6065aa7/attachment.pgp>



More information about the ffmpeg-devel mailing list