[MPlayer-users] mpeg_quant test

D Richard Felker III dalias at aerifal.cx
Thu Nov 6 18:52:45 CET 2003


On Thu, Nov 06, 2003 at 01:21:41PM +0200, Tuukka Toivonen wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> I complained that MPEG-4 made objects moving on dark background leaved
> trails at low bitrates.

Hmm, I thought I was talking about removing ringing or blocking. IMO
there's no good reason why it would remove trails, so sorry if I said
that -- I must not have been reading well.

> Richard suggested using mpeg_quant. I didn't try if
> it helps (I'm using higher bitrates instead), but I did made some PSNR
> tests.
> 
> Result: mpeg_quant decreased hugely PSNR.

This is to be expected. MPEG4 switched to h263 quantizers for a
reason. :)

> Here's the graph:
> http://www.ee.oulu.fi/~tuukkat/mplayer/tests/tt-test-mpegquant/mpegquant.eps
> 
> Here's some info:
> http://www.ee.oulu.fi/~tuukkat/mplayer/tests/tt-test-mpegquant/readme.html
> 
> And here links to other tests:
> http://www.ee.oulu.fi/~tuukkat/mplayer/tests/readme.html
> I added there a link to Colin's tests.
> 
> I didn't add links to On2's nor Corey's VP6 tests, because
> On2 says VP6 is better than H.264 while Corey say libavcodec is better than
> VP6. But certainly H.264 _is_ better than MPEG-4 so something's wrong.

Just because it's a better format doesn't mean the encoder they tested
is better than lavc's mpeg4...

Rich



More information about the MPlayer-users mailing list