[MPlayer-users] Difference between vqscale and vqmin/vqmax values (quality vs. quantizer?)

[D Richard Felker III]

On Wed, Jun 02, 2004 at 06:29:29PM +0300, Jan Knutar wrote:
> 
> > This is nonsense. There's no use for two-pass if quantizer is forced
> > to 3!!
> 
> vcodec=mpeg4:mbd=2:v4mv:trell:cbp:qprd:vqscale=2:psnr (1 pass)
> => Plenty artifacts and errors in resulting bitstream.
> PSNR: Y:46.25, Cb:46.30, Cr:48.92, All:46.60
> 
> 
> vcodec=mpeg4:mbd=2:v4mv:trell:cbp:qprd:vqmin=2:vqmax=2:vbitrate=8000:psnr 
> (2 pass)
> => Ok!
> PSNR: Y:46.25, Cb:46.31, Cr:48.92, All:46.61
> 
> 
> (huh?)
> 
> I'm using a small mpeg2-in-avi clip from a DVD which brings out this 
> subtle little difference (or PEBKAC problem). If needed, i can upload 
> it to mplayerhq...

IMO report the problem on ffmpeg-devel list first and see if Michael
can reproduce it without your clip. If he can't he'll ask you to
upload it.

> That it works with vqmin/max but not vqscale=2 leads me to believe Rich 
> randomly asserting that vqscale might be buggy, to be correct ;-)

Well it wasn't entirely random. When Michael changed the whole lavc
rate control system over to use his lambdas rather than mpeg style
quantizers, the way the "quality" (vqscale) parameter works had to be
recoded too. I suspect it's buggy now.

Anyway let's just get it fixed. Then we can have a good discussion of
xvid vs lavc quality.

Rich