Video shot boundary detection based on MB coding mode and SIFT features on H.264/AVC

摘要

To segment video shot quickly and efficiently, a video shot boundary detection algorithm which combine Macro Block (MB) coding mode with Scale Invariant Feature Transform(SIFT) feature point matching is proposed in this paper. Firstly, the MB coding mode is extracted from H.264/AVC bit stream to calculate the intra coding MB ratio of a frame, according to intra coding MB ratio, some frames are selected out as candidate shot boundary frame; Secondly, the matching ratio of SIFT feature point matching of two consecutive candidate frames is calculated to exclude non-shot boundary frame and retain shot boundary frame. The experimental results indicate that proposed algorithm performs well on both cut and gradual shot boundary detection.