Probabilistic Shot Boundary Detection Using Interframe Histogram Differences

In this paper we study a method for the detection of shot boundaries using interframe histogram differences. Instead of using traditional distance between histograms, we use a probabilistic distance that indicates the chance of a given distance to be a shot boundary. We declare a shot change when the interframe histogram difference is a large deviation from the expected histogram interframe differences given past evidence. Like other histogram based methods the proposed one is very simple while being very robust and effective. The proposed method outperforms similar methods proposed in the literature for the detection of hard cuts and achieves good recall and precision performances for gradual transitions.