So what I need is to break a big video in smaller video by detecting transitions.(I have a small start with this problem)
Then for each smaller video, I have to extract an important(representative) frame, and for this frame I generate a hash based on color histogram, texture or something.
This is the big picture.