[MPlayer-DOCS] CVS: main/DOCS/xml/en mencoder.xml,1.100,1.101

Guillaume POIRIER poirierg at gmail.com
Sun Feb 12 10:23:17 CET 2006


Hi,

On 2/12/06, Loren Merritt <lorenm at u.washington.edu> wrote:
> On Sat, 11 Feb 2006, Guillaume POIRIER wrote:
> > Nico Sabbi wrote:
> >> Guillaume Poirier CVS wrote:
> >>
> >>> CVS change done by Guillaume Poirier CVS
> >>>
> >>> Update of /cvsroot/mplayer/main/DOCS/xml/en
> >>> In directory mail:/var2/tmp/cvs-serv7871/xml/en
> >>>
> >>> Modified Files:
> >>>     mencoder.xml Log Message:
> >>> No sane person should encode last pass without at least 'mbd=2' and
> >>> 'trell' (or should _really_ have _really good_ reasons)
> >>
> >> why trell? My encodes without it look better than with it (generally
> >> fewer artefacts)
> >
> > Well, I assume you know what you are doing and know how to compute your
> > target bitrate so that you don't need trellis. Maybe you just set bitrate to
> > 'inf' ? ;-)
> >
> > However, generally speaking, in cases where you've got bitrate constraints,
> > trell is a big winner, isn't it?
> >
> > Or am I totally wrong here?
>
> Trellis is unrelated to ratecontrol. If it improves compression, then it
> doesn't matter whether bitrate is constrained or not.

Well, that's not what I meant. My understanding of one effect of
Trellis is that it optimizes the coefficients of each macrobloc so
that they will compress better (a sort of bias towards what is known
to compress better), thus saving bits that can be spent elsewhere. So
by its nature, it sorts of introduces errors (i.e. artifacts), but
most of times the bit saving outweigh those errors.
Therefore, given my interpretation of what Trellis does, if your
target bitrate is huge, you may not want to have those errors, and
want to disable Trellis.

Please correct me if I'm wrong, as I don't even know where I heard
this... not in the man page, that's for sure, as that's what it reads
currently:

Trellis searched quantization. This will find the optimal encoding for
each 8x8 block. Trellis searched quantization is quite simply an
optimal quantization in the PSNR versus bitrate sense (As- suming that
there would be no rounding errors in- troduced by the IDCT, which is
obviously not the case.). It simply finds a block for the minimum of
error and lambda*bits.


Guillaume
--
Just because code is syntactically "valid" GNU C doesn't mean gcc can
always compile it.
  Steven Bosscher - 2005-01-01
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11203#c14
---
MPlayer's doc isn't up-to-date. Visit my updated mirror here:
http://tuxrip.free.fr//MPlayer-DOCS-HTML/en/
http://tuxrip.free.fr//MPlayer-DOCS-HTML/fr/




More information about the MPlayer-DOCS mailing list