[x264-devel] Scenecut decision in first Mini-GOP of slice analysis window

[Jason Garrett-Glaser]

On Thu, May 20, 2010 at 5:00 PM, Nageswara Gunupudi
<Nageswara.Gunupudi at harmonicinc.com> wrote:
> Hi,

> 1.  Why is that only the first mini-GOP checked for any scenecuts after
> finding the best number B frames in the analysis window of frames (in
> x264_slicetype_analyse function).

Because if there's a scenecut beyond there, we don't care; it'll be found later.

Also if the scene cut is found, the frame
> is coded as P frame not as I, what the practical reason behind it?

Closed GOP.

> 2.  If any scene change detected in the mini-GOP, the frame-cost analysis
> obtained for the frames after the frame at scenecut are discarded. Is
> understanding correct?

Yes.

> 3.  What exactly the ‘real_scenecut’ argument variable in ‘scenecut’
> function meant for?

If we have frames:

A B C D E F

A scenecut check between A and E is not a real scenecut check.  What
it's checking is whether there's enough of a difference between A and
E that a scenecut might exist between the two.  A real scenecut check
would be between A and B, or D and E -- a check intended to find the
actual location of the scenecut.

Dark Shikari