Adaptive detection of scene change for MPEG videos based on sliding window modeling
基于宏块类型信息的自适应场景变换检测算法

A method for detecting the scene change from MPEG videos automatically was proposed. It used the macroblock type information in the B frame to detect scene changes such as abrupt scene change and gradual scene change. In this method, the sliding window mechanism was adopted to detect the precise location. MPEG-2 was employed as a research platform. The experiment results on movies, news and cartoons illustrate the superiority of the proposed algorithm measured by recall and precision rates.