[XviD-devel] XviD-1.1-Beta 1 bugs summary

Michael Militzer michael at xvid.org
Wed Mar 2 21:28:03 CET 2005


Hi,

I've applied a very trivial fix for cartoon mode. Seems the semantics of a
helper function cartoon mode was using have changed without updating cartoon
mode accordingly. Because of this, cartoon mode was too aggressively forcing
not-coded MBs.

Since I have no test input material to reproduce the fault, please check 
again with the latest cvs version if the problems with cartoon mode are
fixed now.

bye,
Michael


Quoting Michael Militzer <michael at xvid.org>:

> Hi,
> 
> thanks for the links. One thing's for sure: cartoon mode is not intended
> to be used for coding natural images. This will produce artifacts and I
> wouldn't consider this a bug. Regarding the artifacts in cartoon-like
> images,
> it looks as if MBs are skipped that actually shouldn't have been skipped
> (because they're moving). It seems this effect is amplified by either 
> adaptive quantization or b-frames (or both). 
> 
> Cartoon mode is rather aggressively forcing skip mode because cartoons have
> a lot of stationary motion. Wrong skip decisions cause artifacts. I'll try
> to think of something, which eliminates these wrong decisions without
> harming the compression efficiency...
> 
> bye,
> Michael
> 
> 
> Quoting nvidiadx at hushmail.com:
> 
> > 
> > 
> > On Wed, 23 Feb 2005 23:49:30 -0800 Dirk Knop <dknop at stud.uni-
> > goettingen.de> wrote:
> > >Aloha!
> > >
> > >Michael Militzer wrote:
> > >
> > >>Hi,
> > >>
> > >>Quoting Dirk Knop <dknop at stud.uni-goettingen.de>:
> > >>
> > >>  
> > >>
> > >>>- GMC has a bug. With very much motion, some blocks get totally 
> > >>>displaced. The issue is the same no matter if decoded with 
> > >libavcodec or 
> > >>>xvid.
> > >>>    
> > >>>
> > >>
> > >>Do you have a sample demonstrating this effect? Either the coded 
> > >(wrong)
> > >>bitstream or a source file + settings to reproduce the problem?
> > >>  
> > >>
> > >
> > >Someone reporting this bug at doom9 posted an archive here: 
> > >http://s2.yousendit.com/d.aspx?id=1IRU00JPFMEFR26USGL9OUJOF1
> > >
> > >Don't know if it's still online, but it contains the source, the 
> > >avs 
> > >script and some broken encodes ;)
> > >
> > >>>- Some issue with cartoon mode and bvhq (well, cruncher may 
> > >report it 
> > >>>better :) ): some white blocks sometimes floating around within 
> > >high motion.
> > >>>    
> > >>>
> > >>
> > >>Is this really _only_ happening with bvhq? Iirc, the cartoon mode 
> > >doesn't add
> > >>any special optimizations for b-frames - in fact, it wasn't even 
> > >meant for
> > >>using it with b-frames (at least I didn't test it much). Do you 
> > >have a sample
> > >>clip for this problem? I suspect that probably some MBs are 
> > >skipped in 
> > >>b-frames, which shouldn't have been skipped - but you know, 
> > >cartoon mode
> > >>rather aggressively forces the use of skip mode...
> > >>  
> > >>
> > >This is a cruncher-report, he says he can always reproduce it. 
> > >Will ask 
> > >him to report more on it.
> > >
> > >Regards
> > >Koepi
> > 
> > Yes my initial thought was b-vop is couseing this behaviour in use 
> > with cartoon mode but that assumption was wrong it is cartoon mode 
> > + adaptive quantization thats couseing it, other reports and 
> > samples showing that problem can be found @ the following places
> > 
> > http://forum.doom9.org/showthread.php?s=&threadid=90596
> > http://forum.doom9.org/showthread.php?s=&threadid=88367
> > http://cruncher.mufflastig.com/XviD/cartoonbug/
> > 
> > SCNR
> > CruNcher
> > 
> > 
> > 
> > 
> > 
> > Concerned about your privacy? Follow this link to get
> > secure FREE email: http://www.hushmail.com/?l=2
> > 
> > Free, ultra-private instant messaging with Hush Messenger
> > http://www.hushmail.com/services-messenger?l=434
> > 
> > Promote security and make money with the Hushmail Affiliate Program: 
> > http://www.hushmail.com/about-affiliate?l=427
> > 
> > _______________________________________________
> > XviD-devel mailing list
> > XviD-devel at xvid.org
> > http://list.xvid.org/mailman/listinfo/xvid-devel
> > 
> 
> 
> 
> 
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel
> 






More information about the XviD-devel mailing list