Abrupt Cut Detection Based on Motion Information

Shot boundary detection is the first step of content-based video retrieval. In this paper, a novel method of abrupt cut detection is proposed based on motion information for uncompressed video. First we analyze motion vector filed calculated by block-based motion estimation and extract quantitative angle histogram entropy and average magnitude of motion vector filed as metrics of frame differences, which respectively describe distribution regularity degree and intensity of motion vector field. Then design adaptive threshold strategies to identify candidate abrupt cuts, finally eliminate false cuts caused by flashes and gradual transitions using a temporal window. Experimental results show better performance and higher robustness to large camera motions and flashes than histogram-based algorithm.