[Ffmpeg-devel] Bad interation of rc_strategy=1 and qprd

Michael Niedermayer michaelni
Mon Feb 6 02:25:32 CET 2006


Hi

On Sun, Feb 05, 2006 at 10:12:06PM +0100, Nico Sabbi wrote:
> Loren Merritt wrote:
> 
> >On Sat, 4 Feb 2006, Nico Sabbi wrote:
> >
> >>Michael Niedermayer wrote:
> >>
> >>>
> >>>well, you at least have to post 1st pass, 2nd pass stats of lavc, 
> >>>lavc+xvid_rc
> >>>and xvid ...
> >>
> >>
> >>I'm not comparing the two codecs (I prefer lavc much more than xvid):
> >>I'm only encoding to lavc (mpeg2) with and without vrc_strategy=1 to 
> >>see if there is any improvement.
> >>I'll post the psnr figures when I've collected them
> >
> >
> >But you should compare lavc+xvid_rc against xvid+xvid_rc, otherwise 
> >you won't know if you've accurately implemented xvid_rc.
> >
> >First guess for why qprd borks: did you not set lambda appropriately? 
> >xvid's rc deals only with quantizers.
> >
> >--Loren Merritt
> >
> 
> here are the results; if you need more tests let me know.
> Visually mpeg4 looks much better to my eyes, both with vrc_strategy=0
> and with vrc_strategy=1.

hmm, just to make sure ... your xvid version is upto date head cvs?


> PSNR figures for vcodec=xvid are not available.

thats bad, without PSNR stats its hard to know if xvid_rc is buggy or not
you should also run the first pass for xvid_rc with qscale=2, as thats what
xvid does ...

[...]

-- 
Michael





More information about the ffmpeg-devel mailing list