[XviD-devel] Adaptive quantisation

Christoph Lampert xvid-devel@xvid.org
Sun, 4 Aug 2002 19:03:17 +0200 (CEST)


On Sun, 4 Aug 2002, Christoph Lampert wrote:

> > XviD got even disqualified for SPR as the scenes tested were a huge 
> > "block-fest"  :-(
> 
> Do you have a link for that?

Okay, I found it:

Very nice... Let me quote

"Unfortunately I had to disqualify XviD for this test. The results I got
were truly horrible. Blocks all over the place during the first scene, and
the night scene was only half as big as SBC and the impact on quality was
devastating. Since XviD looked much better in the last test I decided to
take the codec out of this comparison rather than to write an extremely
nasty review. "


First of all, I'd say it's _nice_ of him to disquality
XviD. Honestly... screenshots would have been much worse. 

Second, this is exactly what I expected to happen. Boy, I hate this! 
Add some feature, don't test it thoroughly, make it default "because
it's new!!!" and see what happens... you'll have a good reputation ruined
in less than half an hour. 

I would bet that the results would not have been good without luminance
masking! Most likely they would also have been good with old code. 

It's kind of a wonder that the test wasn't a version with buggy b-frames
(but... It a "new!!!" feature) and then XviD would be disqualifie  because
of ugly blocks and missing speed... 


gruel