[MPlayer-users] mjpeg and mencoder

[D Richard Felker III]

On Tue, Jan 07, 2003 at 06:29:10PM -0500, Brad Barnett wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> On Wed, 8 Jan 2003 00:15:09 +0100
> Arpi <arpi at thot.banki.hu> wrote:
> 
> > [Automatic answer: RTFM (read DOCS, FAQ), also read
> > DOCS/bugreports.html] Hi,
> > 
> > > I've been looking for a way to control the quality of mjpeg lavc
> > > encoding with mencoder.  So far, the vbitrate option doesn't seem to
> > > make any significant difference.  If I set it to 600 or to 10000, I'm
> > > just as likely to get the same size file out, and the same image
> > > quality.
> > 
> > hmm. vbitrate should work. anyway you won't get any enjoyable quality
> > under 8000...  even mpeg2 requires at least 8000 for good quality at
> > 720x576
> > 
> > also you can try vqscale=[1..31] for constant quality (but variable
> > bitrate)
> 
> 
> Cool.  I'm trying to basically get as close to lossless compression as I
> can.  I can't do rawrgb, it's just too large.  However, why does something
> like:
> 
> /usr/bin/mencoder -tv on:driver=v4l:width=640:height=480:input=1:norm=ntsc
> -ovc lavc -lavcopts vcodec=mjpeg:vhq:vqscale=1  -oac mp3lame -lameopts
> cbr:br=192 -o /mnt/c/test.avi -endpos 60
> 
> use more cpu than something like:
> 
> /usr/bin/mencoder -tv on:driver=v4l:width=640:height=480:input=1:norm=ntsc
> -ovc lavc -lavcopts vcodec=mjpeg:vhq:vqscale=2  -oac mp3lame -lameopts
> cbr:br=192 -o /mnt/c/test.avi -endpos 60
> 
> Shouldn't it be the other way around?!  After all, I'm doing less
> compression with vqscale=1 than vqscale=2.  Also, what's the best method
> for deinterlacing?  I've tried lavcdeint, but it seems to add artifacts...

AFAIK qscale should not affect performance much either way... Perhaps
it's just the increase in disk io loading down the cpu. You might also
try something like

-lavcopts vcodec=mpeg4:vme=0:vqscale=1

That should get about the same quality, using a good bit less disk
space but slightly more cpu time. You can also add keyint=0 for
intra-only encoding, that'll speed it up a bit at the cost of less
compression.

Rich