Research On Video Shot Boundary Detection Method

Along with the rapid development in computer, multimedia and networking technologies, digital video applications become more and more popular. How to fastly and efficiently retrieve the extremely large amount of video data is the key problem. Content-based video retrieval is developed and becomes the issue of multimedia research.Yet video is not as structuralized as text, this defect prevents the application of video. In order to solve the problem, researchers proposed the solution scheme of "video content structuralizing". This technique is divided into three layers, such as low, medium, high layer of video structuralizing. Shot detection is one of the key technologies of lower layer; shot detection is one of the most important issues in the field of video retrieval, a good shot detection method may be the fundamental of excellent video analysis. Shot boundary detection is always an important topic in digital video processing. Shots are the most basic elements which compose the video sequence. Shot detection provides the foundation for video structure parsing and high semantic level processing.In this paper, we studied fully and systematically on this technology. Main contributions are as follows:Basic concepts, research background and research meanings are introduced. Various shot detection methods in decompressed domain are studied. The traditional algorithms are systemically analyzed, and the issues which are focused are proposed.MPEG video compressed standard is introduced.Ⅰframe, P frame, B frame are studied. It is the basis of shot boundary detection on MPEG video.Various shot detection methods in Mpeg-compress domain are studied. We improve and realize an approach for shot boundary segmentation using MPEG compressed data, it is mainly used histogram and motion vector information, experiment results verifies its effects, but it is still to improve further.