[MPlayer-users] PSNR tests: mbd=2, mv0, trell, and cbp (and other stuff)

Tuukka Toivonen tuukkat at ee.oulu.fi
Mon Oct 13 09:43:16 CEST 2003


On Fri, 10 Oct 2003, Corey Hickey wrote:

>[Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]

Looks like Pine doesn't allow me to quote more than this one line...
anyway...

I made some tests with mencoder/libavcodec using different set of
options and the results are here:

http://www.ee.oulu.fi/~tuukkat/mplayer/tests/tt-test1/readme.html

Warning: the page has eps images. Konqueror at home displays
them correctly (although its ugly as it doesn't antialias),
but in this another computer it crashes it. With Galeon
nor Mozilla it doesn't even try to display the images. So
you may download and view the images separately with "gv"
which allows zooming them anyway [*].

Here are the conclusions from the tests:

>>>
In this investigation, we learned that slightly better quality usually
requires much more computation in MPEG-4. Simply the best options are:

mencoder -mc 0 -noskip -skiplimit 0 -nosound -ovc lavc -lavcopts
vcodec=mpeg4:vhq:trell:mbd=2:vmax_b_frames=1
:v4mv:vb_strategy=0:vlelim=0:vcelim=7:cmp=6:subcmp=6:precmp=6:predia=3:dia=3:vme=4

(except that you should probably use vcelim=0 instead of 7, although it
wasn't tested in this test). Practically equivalent quality is obtained
with:

mencoder -mc 0 -noskip -skiplimit 0 -nosound -ovc lavc -lavcopts
vcodec=mpeg4:vhq:trell:mbd=2:vmax_b_frames=1
:v4mv:vb_strategy=0:vlelim=0:vcelim=7:cmp=2:subcmp=2:precmp=2:predia=1:dia=1:vme=4

(and at very high bit rates even slightly better) but 320% faster. Also
remember to use two-pass encoding if you're using fixed bitrate (vbitrate)
instead of fixed quantization parameter (vqscale).
<<<

That 320% should probably read "76% faster". Sorry.

The scripts used for generating the results are also downloadable.

And thanks for the link
http://fatooh.org/remitests/
Now we can compare these and my results to see if they are consistent.

I have another test running right now, I'll tell you when the
results are available.

I cc this to ffmpeg-devel too since somebody there might be
interested too, I hope.

[*] I consider eps the only sane image format in the web
which is supposed to be device and resolution independent.



More information about the MPlayer-users mailing list