[x264-devel] Re: deinterlacing video prior to x264 conversion

[Tuukka Toivonen]

On Mon, 10 Oct 2005, Andrew Lynch wrote:

>if this is actually doing what I intend for it to do.  I have chosen the
>"-vf pp=md" deinterlace video filter based on a google search.

I wonder if you chose that because I suggested it at some point...?
http://www.ee.oulu.fi/~tuukkat/mplayer/tests/deinttest1/readme.html

md is definitely better than lb as the latter leaves really bad "ghosts",
but nowadays I'm using -vf lavcdeint (ie. -vf pp=fd).
It doesn't leave much ghosts either (just little bit of noise) and
it also doesn't make moving images "rough" which helps compression also.

Also make sure that the material really is interlaced and not
just funnily edited progressive material; in the latter case you should
use -vf phase=T (or something similar) instead of deinterlacing.

>for file in $* 
>do
>  mencoder -ovc x264 -x264encopts pass=1 -nosound -o $file.x264.avi $file
>  mencoder -ovc x264 -x264encopts bitrate=2200:pass=2:8x8dct:i8x8 -vf pp=md
>-oac copy -o $file.x264.avi $file
>done

Hmm, I will try to remember to post my encoding script some day.

PS. In my experiments two-pass encoding doesn't really help compression.
Typically the bit rate is slightly smaller (0.1% or so) and more accurately 
the target rate and the PSNR is practically the same, sometimes slightly
worse even (0.01 dB or so).

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html