[MPlayer-users] PSNR Questions

[D Richard Felker III]

On Sat, Feb 15, 2003 at 09:42:21PM +0100, Struan Bartlett wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> Ian wrote:
> 
> >D Richard Felker III wrote:
> >
> >>How are you measuring the PSNR of the original?? Like I said, you have
> >>to use the same reference to compare PSNR.
> >
> >I've been experimenting with a 3 or 4 pass TV recording method whereby 
> >I capture at maximum bps for a high-quality "original" mpeg4 or mjpeg, 
> >then do a 2- or 3-stage reduction to 1Mb/s mpeg4.  I left the PSNR 
> >option on for both stages and just observed that the figures were 
> >higher for the low-bw output than the max-bw "original".
> >
> >Incidentally, the quality of this procedure was not significantly 
> >(visibly) better than an original grab at the low-bw rate, hence my 
> >interest in an objective measurement. 
> 
> Objective measurement of encoding quality is really my question.
> 
> A week ago a URL (rguyom.ath.cx) was posted of a doc that compared 
> various encoding options by comparing the PSNRs reported by mencoder.
> 
> I was experimenting myself to compare the quality of a cropped movie 
> encoded at bitrate Xbps: first unscaled, but with (mplayer-only) encoded 
> aspect ratio; second, scaled according to encoding-tips.txt's procedure 
> using calcbpp.pl.
> 
> I wanted to use the reported PSNRs to objectively determine the relative 
> performance of each method, but unfortunately it appears this is not a 
> measure that works well.
> 
> 1. What's your experience of the quality of encoded cropped video at a 
> given bitrate in my two scenarios above?
> 
> 2. Does this invalidate the use of PSNR as an objective measure and 
> therefore the results of the rguyom.ath.cx paper?
> 
> 3. If so, how are we to objectively measure the performance of encoding 
> parameters?
> 
> What's your take?

Actually, I just remembered one other good way to 'objectively'
measure quality: average quantizer. Lower is of course better. What's
nice about measuring quality by average quantizer is that the number
should be in some sense proportional to the severity of ringing/block
artifacts.

Using avgq to compare quality between different scaling, however,
probably is not a good idea. At lower resolutions, artifacts become
much more noticable, so a much lower quantizer is needed to make the
video look good. For example, IMO, anything above constant q=2 looks
bad at 320x240, whereas 4-5 is pretty good and even 6-7 is ok (talking
more about peak than average) at 640x???.

But anyway, if you want to evaluate the benefits of deblocking,
deringing, blur, denoise, etc. before encoding, avgq could be a very
useful metric!

Rich