基于宏块类型信息的自适应场景变换检测算法

Adaptive detection of scene change for MPEG videos based on sliding window modeling

提出了一种压缩域中利用B帧的宏块类型信息自适应场景变换检测算法,采取滑窗方式自适应地检测场景突变和场景渐变的准确位置。利用电影、动画、新闻等素材作为实验视频库,并利用MPEG-2作为实验平台,取得了较高的查全率和查准率。

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.