Re: [Mjpeg-users] RC3 aviable

[Florin A. <fl...@an...>]

Bernhard Praschinger wrote:
> 
> In the RC2 there was at high chosen quality factors (-q 1-4) a problem 
> that the quality fell, to a -q factor down to at 10-15.
> That does not happen in the RC3. The Quality factor is lowered, but in a 
> way that makes sense. Dropps from -q 3 to 3.4 - 5.6.

I am under the impression that RC3 has poorer control over the max 
bitrate as reported later by mplex. It's like it either can't get close 
enough to the specified maximum (for -b 8000 -q 8 it only gets to 7400 
for one file, and lower, sometimes much lower, for all other files) or 
it gets too high and overshoots if I start pushing -q just a little bit 
(for -b 8000 -q 7 it's already 9200 on one file, 8800..8900 on several 
other files).

It's hard to tell precisely without doing more tests, but it seemed like 
RC2 and older did not overshoot the max so suddenly with only minor 
changes for -q.

See my previous messages in this thread. Again, the source material is a 
  new but cheap DV camcorder, rather noisy in low light, I'm shooting 
from hand - no tripod, no filtering/processing is done on the video 
stream. I am transcoding to the -f 8 format for DVD. I am trying to 
maximize the image quality while keeping the VOB file parameters "legal" 
(there's only one sound track, AC3 192k, multiplexed with mplex, then 
authored with dvdauthor in a single-title multiple-chapters disk with no 
menus).

A sample of mpeg2enc options is:

mpeg2enc -v 0 -I 1 -f 8 -b 8000 -F 4 -n n -a 3 -o file.m2v -c -q 5 -4 1 
-2 1 -s -D 10 -K kvcd -R 2 -M 2

I only modify -q in these tests.

-- 
Florin Andrei

http://florin.myip.org/