[MPlayer-dev-eng] comprehensive lavcopts test - in progress.

[Roman Gaufman]

Sorry, I dont have the original message, only joined the mailing list 
recently.

Also, you havent answered my question, but I may have phrased it badly.

The question is:

For that long list of opts, if you remove everything, and just run 
mencoder -lavcopts $general:dia=-2:predia-2 
mencoder -lavcopts $general:dia=2:predia2 
($general being psnr:vpass=1, etc, none of the extra opts), do you get the 
same psnr value for the 2 encodings? 

-- if you do, that proves my point of being able to test args with different 
values separately, then combining the ones that give best results with other 
args.

Couple of points:
1) PSNR ratio to filesize is also very important, not just PSNR value, you 
should provide a difference in filesize as well (if any)
2) You get best PSNR values when combining many opts together, yes, but each 
opt can, and should be tested separately to find out what to combine. 

There is no point combing options that show a degradation in quality and a 
gain in filesize. No matter how you group them, they wont help much. So this 
is what I'm doing: I'm testing each opt, to figure out what inpact it has on 
filesize and quality, they I will combine. 

I understand the gain in quality/ loss in filesize wont remain consistent when 
I start grouping those args together, but it will lower the number of 
combinations needed to test dramatically.


On Monday 14 June 2004 16:31, Romain Dolbeau wrote:
> Roman Gaufman wrote:
> > Cant say I see your point.
> >
> > From your results I can tell that
> > [dia=-2,predia=-2] and [dia=2,predia=2]
> >
> > or:
> > [dia=-1,predia=-1] and [dia=2,predia=2]
> >
> > dont have an PSNR change.
> >
> > Do you have tests for them individually without all the other args?
> > Wouldn't they give you the same findings?
>
> The particular cases mentioned in that mail are an answer
> to Michael's question wrt local maxima of PSNR. They're
> an extremly small subset of the result.
>
> > Also, what type of file are you testing, PSNR value is very different
> > with different bitrate, with different input type, like animation or
> > video, slow motion or high motion.
>
> As mentioned in my original message, it's the beginning
> of the european DVD of "Crouching Tiger, Hidden Dragon".
> The bitrate and common options are also in the original
> message.