A Novel Shot Edge Detection Algorithm based on Chi-square Histogram and Macro-block Statistics

A rapid approach to shot boundary detection was introduced, which was based on the DC histogram Chi-square test and macroblock type statistics in one frame of MPEG-2 video sequence. Some information about DC histogram differences between I-frames was extracted first to detect in which GOP (group of picture) there are shot boundaries through the threshold active training method proposed in this paper. Then the similarity and similarity difference between frames of the GOP was calculated to determine whether it is cut shot or gradual shot. The experiment results show that this method had a high accurate rate in both cut shot detection and gradual shot detection. Furthermore, there is no need to decompress the MPEG-2 video sequence, so the speed of computing is higher, and the computational complexity is smaller.