Shot boundary detection based on the knowledge of information theory

Shot boundary detection servers as the preliminary step to structure the contents of videos. Many research efforts have been devoted to this field. The distributions of this paper are as follows. First, in contrast with former methods, shot boundary detection is performed based on the knowledge of information theory; the threshold is determined automatically. We find that mutual information, joint entropy and conditional entropy all can detect the abrupt changes. Second, we also analyze the gradual transitions and motions. Fade in/out is easy to be detected using the joint entropy. Since the dissolve and motion have different characters in our method, it is easy to distinguish them from each other. The experiments based on the proposed method show better results and further researching is worthwhile.