Research On Shot Boundary Detection In Compressed Domain

With the rapid development in computer, multimedia and network 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. As a result the content-based video retrieval (CBVR) technology emerges and becomes the issue of multimedia research.Shot boundary detection is the most fundamental operation and a very important step in content-based video retrieval. According to the relevance of adjacent frames, it divides video data into physical units-shot. The detection precision directly affects the retrieval results of CBVR system. So the accuracy and the speed of shot boundary detection algorithm are critical.Now, a majority of video has been deposited into compressed format, so the shot boundary detection in compressed domain is the focus of the research.Firstly, a brief introduction about the basic theory of video compression and MPEG-2 video compression standard is described; secondly, shot boundary detection algorithms in uncompressed domain are introduced; thirdly, a shot boundary detection algorithm based on Macroblock type is presented and another algorithm base on HSV color histogram and automatic threshold is proposed. Furthermore, the sliding window algorithm is used to increase the detection precision. The results of experiment show our algorithms realizy the shot boundary detection from long video sequence and have good detection precision, and can be applied to various videos.