[MPlayer-users] Re: encoding quality - bpp, resolution, quants and quality

[HR]

> HR wrote:
> > As the codecs get better and better, how true is the bpp 
> recommandation
> > table in encoding-tips.txt now and in the near future?
> 
> Those are just intended to be rules of thumb. I've encoded 
> some stuff at 
> less than 0.2 BPP which looked great and some stuff at 
> greater than 0.25 
> BPP that still had very noticeable artifacts. The minimum 
> acceptable BPP 
> varies hugely with quality and type of source material, quality of 
> decoder and screen, and especially with the observer.
Yes, indeed. So more often than not, I end up doing several encodings to
squeeze the best out of the bits. Does anyone have any experience with
the XViD mod_quant? It seems like a good idea, but is it really? And I'm
assuming that there's no reason not to use the default me_quality value
of 6?

> > quant of 2. I should propably scale it up a bit then, but 
> how can it be
> > that ALL the quants = 2?
> 
> That does seem odd. Was the resulting BPP really 0.199? In 
> the end, all 
> that matters is perception. Try a larger frame size and see 
> if it still 
> looks OK. Video encoders are still pretty stupid. Only a 
Well, my script can ofcourse be buggy, but I'm using bpp (or actually,
bpb, since that's a tad nicer when doing non-float math) as an input
value to find the best resolution, also considering the AR. So if the
resulting bpp isn't what I asked for, it's the decoder to blame given my
script does its math right. I can supply the resolution and bitrate and
such for someone to verify the sanity of my script... Trying a slightly
increased resolution now for a bpp of 0.171, but there's already smoke
coming out of my Athlon encoder "slave", so it'll be a few hours... ;)
If the quants still are all 2 then I don't know... (no, I'm actually
*not* using fixed_quant=2 or anything like that...)

HR