[MPlayer-users] mencoder cropping does not work?
D Richard Felker III
dalias at aerifal.cx
Sun Jan 11 05:12:34 CET 2004
On Sun, Jan 11, 2004 at 05:54:44AM +0200, Jan Knutar wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> > > -ovc lavc -lavcopts vcodec=mpeg4:mbd=2:v4mv:vqscale=2
> > Why not qscale=1?
> Sure, but only #ifdef _PURIST :)
> > > better quality, but slower:
> > >
> > > -ovc lavc -lavcopts vcodec=mepg4:mbd=2:v4mv:vqscale=2:trell
> > As I understand it, this would be "lower filesize, but slower" and
> > not "better quality, but slower".
> It's both, according to psnr values :)
> reencoding with above options some short avi to mpeg4
> Without trell:
> ~25fps, 1245.206 kbit/s, PSNR 48.14
> With trell:
> ~16fps, 1233.866 kbit/s, PSNR 48.69
> let's test some more...
> ~42fps, 1252.492 kbit/s, PSNR 47.82
> the same, with vqscale=1
> ~41fps, 1252.492 kbit/s, PSNR 47.82 (hm, that's strange)
> ~25fps, 1336.927 kbit/s, PSNR 48.40
> ~16fps, 1419.456 kbit/s, PSNR 49.41 (trell increases both filesize and
> quality with qscale=1?)
> vqscale=1 didn't behave the way I expected :\ Didn't think it would be
> such a small bitrate increase, and no difference at all without mbd and
Hmm, I guess qscale=1 gets rejected without mbd=1 or 2; if I remember
correctly, certain macroblocks cannot be encoded correctly with q=1.
If would be nice to see the real explanation of this, though.
More information about the MPlayer-users