[x264-devel] Adaptive B-Frame decision?!

[Peter.List at telekom.de]

Hallo Savvas,
the explanation from Jason seems to say, that the irregularities 
only appear at the very end of the GoP and are only a fallback to
P, P, P, P,
I think I could deal with that....

But maybe you should reduce --scenecut 40 to a lower value, because
we had seen much to many scenecuts?

Regards, Peter



-----Ursprüngliche Nachricht-----
Von: x264-devel-bounces at videolan.org [mailto:x264-devel-bounces at videolan.org] Im Auftrag von Jason Garrett-Glaser
Gesendet: Mittwoch, 2. Februar 2011 20:01
An: Mailing list for x264 developers
Betreff: Re: [x264-devel] Adaptive B-Frame decision?!

On Wed, Feb 2, 2011 at 8:38 AM,  <Peter.List at telekom.de> wrote:
> Hallo Jason,
>
> thank you for the quick and conclusive reply.
> I guess scene cut detection would still work with --scenecut 0
> but has the lowest sensitivity, or is it effectively turned off
> (without using the --no-scenecut command)
>
> I am curious, why there is this connection between the two commands
> "--b-adapt 0" and "--scenecut 0"

--scenecut 0 is off.

The connection is because x264 will not place B-frames across a point
where scenecut detection triggered.  So, for example:

A B C D

Suppose x264 wants to do:

P b b P

But suppose scenecut detection says "D" is different enough from "A"
to warrant a scenecut.  Then, x264 can't do the above.  However, it's
possible that no single frame difference is enough to trigger
scenecut, so a keyframe is never actually placed.  This can happen in
short areas of extreme motion.

Jason
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel