一种基于宏块类型匹配的镜头切换检测算法

提出了一种新的MPEG压缩视频镜头切换检测算法。该方法通过分析切换点附近各帧的宏块类型特征，引入了一种新的“宏块类型匹配符”（MTMO），使用其在宏块级别上度量子图片组（SGOP）内部各帧之间的差异性，并结合一种自适应阈值选取技术，完成镜头切换检测。实验结果表明，所提算法适用于不同SGOP结构的MPEG视频序列，查全率达99％，准确率达95％。

Shot Cut Detection Algorithm Based on Macroblock Type Matching Information

A novel shot cut detection algorithm for MPEG coded videos is developed.The proposed method exploits the specific characteristics of the macroblock type information for adjacent frames when a shot cut occurs.A new macroblock type matching operator(MTMO) is introduced to estimate the dissimilarity of the consecutive frames of sub-group-of-picture(SGOP) in the macroblock level.An adaptive threshold selection scheme is used to identify the location of shot cuts. Experimental results on MPEG streams with different SGOP structures show that the proposed algorithm can detect shot cuts with 99% recall rate and 95% precision rate.