[MPlayer-users] lavc psnr mixup

[Meino Christian Cramer]

From: D Richard Felker III <dalias at aerifal.cx>
Subject: Re: [MPlayer-users] lavc psnr mixup
Date: Wed, 8 Sep 2004 02:01:03 -0400

Hi,

 does all this discussion of psnr (which I do not understand fully)
 mean, that the PSNR output is simply spoke wrong and the assumption
 that it has anything to do with the quality of the encoding result is
 also wrong ?

 Slightly confused,
 Meino


> On Tue, Sep 07, 2004 at 10:28:23PM -0700, Loren Merritt wrote:
> > ... That said, while I do find that B-frames improve both psnr and
> > visual quality in most real movies, I have come to agree that they still
> > suck and should be replaced by some other 2nd order temporal prediction
> > (3D wavelets?).
> 
> 3D wavelets are just about the worst idea ever introduced in video
> encoding. ;) Video is horribly undersampled in the temporal dimension,
> so any sort of linear transformation in time is guaranteed to be
> bogus. A few of us had a hypothesis a while back that you could first
> compute motion paths through a collection of frames, and then do a 3D
> wavelet transform along that motion-warped time axis, and benefit from
> it, but a theory like this remains to be tested.
> 
> Rich
> 
> _______________________________________________
> MPlayer-users mailing list
> MPlayer-users at mplayerhq.hu
> http://mplayerhq.hu/mailman/listinfo/mplayer-users
>