By "hashing" all frames in some sensible way (and ignoring blank frames) it should be possible to find a series of repeated frames that are near identical for a second or so. These blocks of identical frames would then be our commercial lead in and our markers (perhaps optionally looking for a scene or blank frame afterwards to find the exact end of the commercial break.
Any thoughts on a sensible hashing function? Might be almost a perfect identical frame for DVB users, but for analogue users there is going to be sufficient "noise" to make the detection of identical frames tricky. My first thought would be to break the frame up into, say, 8x8 pixel blocks, average the colour and then quantise this colour to a smaller range (say 1-8 instead of 1-256 for each colour). Does this sound "fuzzy" enough to spot identical frames? Anyone read any research papers on doing this that might have some better ideas..? Anyone got a couple of hours free to try coding this up...?
Ed W _______________________________________________ mythtv-dev mailing list mythtv-dev@mythtv.org http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-dev
