[MEncoder-users] new doom9 codec comparission (submission)

Corey Hickey bugfood-ml at fatooh.org
Thu Dec 15 04:52:34 CET 2005


Oops, I forgot to send this to the list.

Corey Hickey wrote:
> Doom9 Feedback Hotline wrote:
> 
>>Alright, the results are on. Good news and bad news for you.
>>
>>The good news is that the mpeg4 codec qualified. Initially I found no 
>>difference to XviD but as the review progressed it was made clear that the 
>>main round won't be a piece of cake. From all the ASP codecs in the 
>>qualification round, XviD performed the best (except for speed but you can 
>>read that once the article goes online).
>>
>>As far as Snow goes, I'm afraid I encountered some problems. Overall quality 
>>impression is not bad, but MPC constantly crashing is worrysome at least. 
>>However, the final decision was not made because of MPC (it was known 
>>upfront though.. MPC is a must every participant has to go through, it got 
>>another into bad trouble - a codec that would be disqualified prior to 
>>review), but because of a lingering impression I got of tiny spots of 
>>discoloration.  I experienced the same, bug on a massive scale, with the 
>>only other wavelet codec I have tested: Dirac. It's quite hard to put the 
>>finger on it but I think I managed to catch just the scene. I'm attaching 3 
>>screenshots, one from the source, one from Snow, and one where I marked the 
>>areas affected with red. I also extracted a few GOP interval's worth of the 
>>AVI, as well as the corresponding XviD stream. Not that every codec has a 
>>tendency towards green with certain facial details, but it's so strong that 
>>it raises your eye in Snow.
> 
> 
> Yeah, that does happen sometimes. I'm temporarily mirroring the
> attachments (with zip file extracted) at:
> http://fatooh.org/files/doom9tmp/attach/
> 
> 
>>Last but not least, I initially started reviewing with the VMR9 renderless 
>>renderer in MPC, but found that both Theora and Snow (incidentally both 
>>decoded via ffdshow, where the rest of the codecs was decoded by ateme's 
>>filters, plus AviSynth is decoded by the Helix YV12 codec) were discolored, 
>>too bright and washed out. That didn't happen when I changed the renderer to 
>>the system default (which is useless when it comes to taking screenshots.. 
>>good thing there's a VfW codec for these two and no postprocessing in the 
>>playback filter).
>>
>>So in the end with Snow we have slight picture problems and some issues on 
>>the usability side. I'm afraid the qualification round is just there to weed 
>>those issues out, but I'm confident that Snow will make it the next time 
>>round if some time is invested to solve these issues (and ratecontrol).
> 
> 
> Ok. Too bad. Thanks for giving it a try.
> 
> -Corey
> 




More information about the MEncoder-users mailing list