[MPlayer-users] Re: variable bitrate, fixed quality

D Richard Felker III dalias at aerifal.cx
Thu Jul 10 19:48:55 CEST 2003


On Thu, Jul 10, 2003 at 03:16:49PM +0200, Nikolaus Rath wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> Jonathan Rogers <jonner at teegra.net> wrote:
> > Nikolaus Rath wrote:
> >> Of course, this makes sense. But unfortunately the manpage doesn't say
> >> what the rate control function actually is and what it can return. So
> >> even if you understand the function you won't have the slightest idea
> >> in which way this modifies the codecs behaviour.
> > 
> > The default function actually is "tex^qComp" according to the
> > manpage. Read it, you'll like it.
> 
> I've read it, really.
> 
> >> Maybe you can say a bit about the return value of the function? How
> >> does the rate controlling work?
> > 
> > I don't know the all the details, but the manpage does say that if the
> > function has a value of 1, bitrate will be constant and if it has a
> > value of "tex" (texture complexity) there will be constant
> > quality. Clearly, the closer qComp (or vqcomp) is to 1, the more
> > constant the quality will be, since n^1=n.
> 
> Yes, that's clear. But what kind of value is "tex" Is it in a certain
> range (0 - 1)? What happens if I set vrq_eq=0.5? What the manpage
> actually says is that a return value of 1 means constant bitrate and
> some other value ("tex") means constant quality. Nothing more. You
> can't really customize the function with these informations. What is
> the allowed range (and scale) of return values and how do they
> influence encoding? What scale and in what range are the mentioned
> variables? 

>From what I've been told, the range of the function is totally
irrelevant. vrq_eq=0.5 should behave identically to vrc_eq=1, or any
other constant. What matters is how it varies with respect to
complexity and whatnot.

However, having read the rate control code, I found it all very
confusing, and I really have no idea how it's supposed to do that...
So perhaps my understanding is wrong, or perhaps the code is just
weird or buggy or something. :)

Rich



More information about the MPlayer-users mailing list