ronnylov
4th February 2011, 14:53
Hi!
I read an interesting article regarding how to detect TV commercial breaks:
http://www.ic.uff.br/iwssip2010/Proceedings/nav/papers/paper_126.pdf
The luminance and the
two chroma components of the video frame are divided
into M×M blocks. The differences between a given
block’s DC value and its neighboring block’s DC values
are used to form the frame’s visual hash. The use of DC
values provides the invariance of the computed hash to
high frequency modifications.
Can this be done in avisynth alone, like creating a very small resolution video made by the fingerprints which then can be stored as a lossless avi file?
I had one idea that I could use DCT Filter and somehow measure a value in each block to calculate a new pixel for the fingerprint. But then it seems to use differencies between neighbouring blocks also.
This is some kind of test I want to do, related to this thread:
http://forum.doom9.org/showthread.php?t=159079
I read an interesting article regarding how to detect TV commercial breaks:
http://www.ic.uff.br/iwssip2010/Proceedings/nav/papers/paper_126.pdf
The luminance and the
two chroma components of the video frame are divided
into M×M blocks. The differences between a given
block’s DC value and its neighboring block’s DC values
are used to form the frame’s visual hash. The use of DC
values provides the invariance of the computed hash to
high frequency modifications.
Can this be done in avisynth alone, like creating a very small resolution video made by the fingerprints which then can be stored as a lossless avi file?
I had one idea that I could use DCT Filter and somehow measure a value in each block to calculate a new pixel for the fingerprint. But then it seems to use differencies between neighbouring blocks also.
This is some kind of test I want to do, related to this thread:
http://forum.doom9.org/showthread.php?t=159079