[MPlayer-users] Quality suffering with lavc
Michael Niedermayer
michaelni at gmx.at
Sun May 11 02:17:17 CEST 2003
Hi
On Saturday 10 May 2003 11:28, Nico wrote:
[...]
> mencoder -nosound -ovc lavc -lavcopts
> vpass=1:vcodec=mpeg1video:vhq=1:vbitrate=$BITRATE:keyint=16:vrc_maxrate=
> 9500:vqblur=0.0:vlelim=-4:vcelim=-9:dark_mask=0.0:naq:vqcomp=0.7:vqmax=7:mb
>qmax=12:vqdiff=2:cmp=0:subcmp=0:mbc mp=0:trell -of mpeg -sws 2 -vop $EXTRA
> $INFILE -ofps 25 -o $OUTFILE
[...]
> The result is usually impressive, but dark scenese are always very ugly,
> because dark/black area are always greyish and full of blocks.
> I uploaded this file to mphq to show the problem:
> ftp://ftp.mplayerhq.hu/MPlayer/incoming/BlockyBlack/harry.mpg
[...]
> How can I remove (or at least reduce) this effect?
could u try to reduce the options used to the minimum needed to cause the
effect, so that removial of any option left fixes the problem (could also be
that no options are needed at all to cause it ...)
>
> Another strange effect that I noticed is that when NOT resizing
> (whatever the dimensions) the quality is always much higher then when
> shrinking (with -sws 2), although theoretically the bpp rises.
hmm, i can imagine several reasons why, alltough i dunno if any of them is
responsible for it
1. -sws 2 is over sharpening the image so its more difficult to compress (try
-sws 1 or try -sws 2 -vop scale=w:h:0:X with different X)
2. as the source is probably MPEG1/2/4 which is 8x8 DCT/MC based its quite
easy to represent this accurately again as 8x8 DCT/MC but if its scaled then
it gets more difficult ...
btw, how did u measure quality? double blind viewing test or psnr ;)
[...]
--
Michael
screen[y][x] ^= cursor[cy][cx]; (violates patent #4,197,590)
median(mv[y-1][x], mv[y][x-1], mv[y+1][x+1]); (violates patent #5,905,535)
buf[i]= qp - buf[i-1]; (violates patent #?)
for more examples, see http://mplayerhq.hu/~michael/patent.html
stop it, see http://petition.eurolinux.org & http://petition.ffii.org/eubsa/en
More information about the MPlayer-users
mailing list