[MPlayer-users] black squares in encoded film

[gaddo]

Loren Merritt wrote:

>On Wed, 12 May 2004, gaddo wrote:
>
>  
>
>>I am seeing black squares in my encoding from dvd with mencoder.
>>(please see attacched image to see an example). The black squares
>>should be white ...
>>
>>here is part of my encoding script ..
>>
>>LAVCOPTS='vcodec=mpeg4:vbitrate=1350:mbd=2:vqmin=1:lmin=1:v4mv:pbias=-16:cmp=2:subcmp=2:psnr:vmax_b_frames=1:acodec=ac3:abitrate=224:trell:cbp'
>>
>>mencoder -o test-p1.avi dvd://$TITLENUM -dvd-device $DVD -vf
>>'crop=720:424:0:76,scale=640:288' -ovc lavc -lavcopts
>>$LAVCOPTS:'vpass=1' -oac lavc -channels 6 -alang it
>>mencoder -o test-p2.avi dvd://$TITLENUM -dvd-device $DVD -vf
>>'crop=720:424:0:76,scale=640:288' -ovc lavc -lavcopts
>>$LAVCOPTS:'vpass=2' -oac lavc -channels 6 -alang it
>>
>>i have tryed to isolate the minimus set of options and the minimum slice of
>>film but is hard to do because the effect is somewhat related to the
>>bitrate and the complexity of the images; I have seen the effect
>>without ac3, cbp and lmin=1.
>>
>>in the first pass the images are _very_  blocky  but i can't  see the black
>>squares.
>>    
>>
><snip logs>
>
>I saw something like that in one of my encodes.
>The movie was mostly pretty low motion, but had a few very complex scenes.
>When I used vbitrate=x on both passes, the first pass used quant 1 or 2
>for most of the movie, but the complex scenes got quants around 10-31.
>Then on the second pass, the low motion scenes got quants 3-4 and the
>high motion got 1-2, with lots of those black blocks.
>
>  
>
Hello
i can clearly see the problem at the end of complex scenes
and i saw black squares only when the image was very
bright. This seem like to be an overflow problem with
big quantizers or something like it. Anyway seem to
be a bug somewhere. Or maybe i didn't see black
blocks on a dark background ...

>I solved it by running the first pass with vqscale=2. This produces better
>quality than bitrate-based first pass in all cases I've seen, even when I
>don't experience such extreme problems.
>(And vqscale=2 works slightly better than vqscale=3, even if the second
>pass results in average quant around 3 or 4)
>
>--Loren Merritt
>
>  
>
Interesting... How do you judge the quality? By psnr or
looking at the encode? How much the psnr is different
between vqscale and vbitrate ?

ciao, marco.

>_______________________________________________
>MPlayer-users mailing list
>MPlayer-users at mplayerhq.hu
>http://mplayerhq.hu/mailman/listinfo/mplayer-users
>
>
>  
>