An adaptive cumulation algorithm for video shot detection

Abstract

A cumulation algorithm for video shot detection is first introduced. The algorithm considers the properties of gradual transitions. In a gradual transition, there is only a small difference between consecutive frames. The algorithm can remember the differences between consecutive frames and cumulates them. When the accumulation difference exceeds a threshold, an occurrence of shot transition is declared. Because each movie has its own color and light properties, we consider the color mean value as an element for deciding an adaptive threshold. It was observed that the saturation value drops during the dissolve process. Therefore, we consider the saturation value too in order to estimate the adaptive threshold. For dissolve and fade, we make full use of their special features to further improve the performance of the algorithm. We tested our algorithm with clips extracted from MPEG video CDs. The algorithm showed good performance in detecting the gradual transitions as well as the abrupt cuts.