[x264-devel] VP8 versus x264

[Jason Garrett-Glaser]

On Mon, Jun 7, 2010 at 1:44 AM, Klaue, Jirka <Jirka.Klaue at eads.net> wrote:
>>> I've compared VP8 with x264, just PSNR and SSIM so far.
>>> If you are interested: <http://www.tkn.tu-
>> berlin.de/research/evalvid/EvalVid/vp8_versus_x264.html>
>
>> If you're doing metric comparisons, turn psy opts off in both
>> encoders.  That means x264, too.
>>
>> Use --tune psnr or --tune ssim accordingly.  Note that VP8 by default
>> is tuned for PSNR.
>
> I'll add the results using <http://www.its.bldrdoc.gov/vqm/> soon in the
> hope that
> it will be more meaningful than statements like "the picture looks
> better". It should,
> since the metrics NTIA used match well to tests with humans.
>
> BTW, do you have an example of a video which looks "better" with a lower
> PSNR?
> And, how do you measure this?

Easy: encode any video with --tune psnr and without --tune psnr.
It'll look much better without --tune psnr, but will have much lower
PSNR.

Dark Shikari