[MEncoder-users] capture: synchronization problems

[RC]

On Wed, 23 Mar 2005 12:47:54 +0200
Oded Shimon <ods15 at ods15.dyndns.org> wrote:

> You should capture using some kind of high quality (or even lossless) codec

Yes, MPEG2/4 are both very high quality (lossy) codecs...  What's the problem? 

Capturing losslessly is a good idea if you have enough space, and want to squeeze every last bit of quality out of it, but at very high bitrates, the difference is minimal.

> and then re-encode properly to whatever 
> target codec you intend.

Yes, that is the intention.  Using a completely different lossy codec is not a good idea though...  You lose different bits of visual information from each different lossy codec, making the final result look worse, and for no good reason.

> If you set keyint to 1, just about the only gain made by using these codecs is 
> lost.

No, size is not the only gain, because with keyint=1, they are FASTER than mjpeg.  Plus, there is the question of losing quality.

> I don't think you'll get any faster/better results with other codecs.

I've already shown MPEG2/4 are faster.  Better is debatable.

Let's try another test, shall we?

MPEG4
5981.221 kbit/s
PSNR: Y:44.18, Cb:47.15, Cr:48.22, All:45.06

MPEG2
6663.826 kbit/s
PSNR: Y:44.25, Cb:47.89, Cr:49.08, All:45.25

MJPEG
7524.783 kbit/s
PSNR: Y:44.75, Cb:48.25, Cr:49.41, All:45.73

Unfortunately, I couldn't get all three to use the same bitrate (might take lots of trial-and-error--anyone know how to improve singlepass ratecontrol?), but MPEG2 and MPEG4 are providing nearly as good a PSNR number as the MJPEG encode, but at MUCH lower bitrates than MJPEG.

> Feel free to flame me, but only if you have 
> proof behind your flames.

Why?  Sounds rather hypocritical, as you haven't provided one bit of proof for any of your claims.