[mythtv] Another commercial flagging method

[Chris Pinkham]

> I've been experimenting with commercial detection some more and I think I 
> have a pretty good method now. It looks for suspicious events in the form of 
> blackframe scenes of a standard length and scenechanges at ten second 

<snip...>

> within two frames for the ten-second test and the framrate doesn't fit well 
> with integers. Any thrashing is appreciated, the patch needs to be applied 
> to fresh cvs.
> 
> Liam

I saw your email about NTSC and am assuming I apply the first patch in that
email then the 2nd.  I was looking through the code earlier tonight but
haven't had a chance to apply the patch(es) and give them any testing.
Looks to be be better and have some more logic than the scene change code I
had originally put in (which doesn't work too well I might add).

I'll try to get this put into my tree and give it a runthrough with some of
the shows I have recorded.  If it doesn't negatively affect any of the
existing detection then I can probably commit this to CVS after v0.14 is
released so it can get a wider testing base.  If this is better than the
existing scene change stuff (which is probably is) then we could take out
the current scene change code before v0.15.

How well does it appear to be working for you and I'm assuming the current
blank-frame only detection doesn't work so well, correct?

I have some preliminary logo detection code which works pretty well.  I was
playing with it late last year.  It does a pretty good job at detecting
logos but I never got to the point of using the detected information to
try to determine commercial start/stop points.  I'd like to end up with
a detection "method" that takes into account all other detection methods,
so it would use blank frame, scene change, logo, audio, etc. to try to be
as accurate as possible.

-- 

Chris