[MPlayer-users] Should vqscale be used for first pass?

[Attila Kinali]

On Sun, 29 Jun 2003 08:32:35 -0500
Jonathan Rogers <jonner at teegra.net> wrote:

 
> Do you know why the manpage advocates using vqscale for the first of two 
> passes? Obviously, this isn't how you do it, since that would make 
> countquant.pl useless. When I've noticed PSNRs for the second pass, it's 
> seemed that using vqscale for the first pass didn't do as well as using 
> just vbitrate. Perhaps the manpage is misleading and should be changed. 
> I think I've read that XviD does use a constant quantizer for the first 
> pass, so perhaps that's where the confusion originated. Does anyone have 
> an authoritative answer? A lot of lavcopts are still magic to me. I 
> guess I need to start reading the source.

I'm not an expert on this field but afaik it comes from that you have
to rate the images to have a proper value on which you can decide
which number of bits you can use for a particulary image.
Now, if you use vbr for the first pas, you get "size" values that depend
on other frames too.
But, this works only if the value used for vqscale is more or less about
the actual encoded one. So, if you use a value that is far away from
that what actualy will be the bit rate per pixel you'll get better
values with vbitrate as it's bit/pixel is nearer to the actualy encoded.

I hope this is right, if not... dont shot me :)

			Attila Kinali

-- 
Emacs ist für mich kein Editor. Für mich ist das genau das gleiche, als wenn
ich nach einem Fahrrad (für die Sonntagbrötchen) frage und einen pangalaktischen
Raumkreuzer mit 10 km Gesamtlänge bekomme. Ich weiß nicht, was ich damit soll.
		-- Frank Klemm, de.comp.os.unix.discussion