视频镜头边界检测系统的设计与实现

With the rapid development of information technology, more and more video data are being produced, and video query and retrieval is becoming an important research subject. Shot is a basic unit of video. Video needs to be broken down into shots so that video content can be effectively organized and accessed to. Therefore, video shot boundary detection is the basis of video analysis and retrieval. This paper provides a shot boundary detection scheme based on support vector machine (SVM). The scheme adopts SVM classifiers as the main method of shot boundary detection by extracting features from decoded MPEG video data. In the scheme, image edge detector and motion detector are designed to eliminate adverse affects to shot boundary detection caused by flashlight, video-in-video and motion. The testing results show that this scheme performs well in shot boundary detection.