Re: [rtcweb] Performance of H.264...

[John Leslie <[e-mail]>]

Maik Merten <[e-mail]> wrote:
> Am 20.11.2013 18:52, schrieb Bernard Aboba:
> 
>> http://iphome.hhi.de/marpe/download/Performance_HEVC_VP9_X264_PCS_2013_preprint.pdf
> 
> There's much to be said about PSNR as means for comparing image quality. 
> There's also much to be said about PSNR as means for conducting 
> cross-codec quality assessments.
> 
> And little of it is pretty.

   I don't think we need to pick on the authors here -- they were quite
clear about what they were doing.

   IMHO, the point to take away from the paper is that _neither_ H.264
nor VP8 are considered "current" first choices.

   The paper basically showed _by_how_much_ H.264 falls short of current
state-of-the-art.

> I always chuckle when I see x264 being given the "--tune psnr" 
> parameter. Even core developers of x264 have strong opinions on this: 
> http://x264dev.multimedia.cx/archives/458

   Fun reading...

   But we _don't_ need any more lengthy criticisms of comparisons right
now. We _aren't_ going to drop everything and switch to H.265 as MTI --
or any other state-of-the-art codec.

   I fully agree that both H.264 and VP8 deserve SHOULD status; and
I agree H.261, being good enough for sign-language reading, looks like
the right fallback.

   H.261 is certainly easy enough to "implement" and deploy; and I'll
bet 80%  of us are ready for the question...

--
John Leslie <[e-mail]>