[MPlayer-users] RFC: docs update for "how to create a high quality DVD rip"

[Wayde Milas]

On Mon, 2004-06-07 at 23:30, D Richard Felker III wrote:

> > Well, you dont want every frame to be q=2 do you? I was under the
> > assumtion that you can get away wtih q > 2 on low motion frames.
> 
> No, the opposite. You want q=2 (or even 1) on low motion frames where
> even minor artifacts will be visible. On high motion, 3 or 4 is fine.
> 
> > Let me ask this. Does q=2 use as much space on a fast motion frame as it
> > does on a low motion frame? I always thought solid q values used about
> > teh same bits no matter what they were encoding?
> 
> Nope, totally backwards. q=2 means you keep lots of precision, so it
> uses however much space is needed to keep all that precision.

Wow.. I should have known that. I just never thought about it the right
way. This makes more sense.

I wish there was a way to feed wider parameters into mplayer instead of
a solid  q or a set bitrate. More like use a really low Q on slow motion
frames, higher Q on fast motion, don't go over a max Q, ect.

I see vmin and vmax, but that doesnt really help much though in
practical terms?
> > Encode a 240 second chunk at various bitrates and graph it. PSNR values
> > are not linear. Its that non linearity that has me pondering different
> > schemes.
> 
> The PSNR values you see measure error logarithmically.

Duh. Just like sound. I knew that. I swear. :)