On Wed, 2006-10-25 at 12:37 +0200, Duncan Webb wrote:
> There is an interesting paper, not about ads but about finding cuts in
> news streams, possible the same system could be used to find ads.
>
> http://www2006.org/programme/files/pdf/p52.pdf
Did you just get back on topic? I'm all confused.
:)
I skimmed the paper but it seemed pretty light on details. I haven't
really been following the approaches for commercial detection, but I
always assumed I would implement it like this:
* Mark video file into potential commercial boundaries, using:
* Black (dark) frame detection
* Sudden scene changes accompanied by some noticeable
difference in volume. If we use a 2-pass approach, we
can compute the median volume and then mark amplitude
variances (with some threshold to be determined
empirically) from the median.
* Where the space between two boundaries is called a gap, look for
groups of gaps where each gap is fairly close to 15, 30, or 60
seconds. If there are 3 or more of these gaps, then assume a
block of commercials.
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Freevo-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freevo-users
