The Study Of Algorithm Of Shot Boundary Detection Including Camera Motion In MPEG Compressed Domain

When the digital information age is coming, people are very convenient to access a large number of video data. The organization, summarization and retrieval of video data have become the focus research. The Content-Based Video Retrieval(CBVR)system is a good solution to this problem, and it has being a hot topic in multimedia field for recent years. The first step to index and search the un-structural video stream is to segment them into basic unit-shots. Using shot as basic unit, video content can be extracted, video index and retrieval scheme can be performed. Shot boundary detection is a key technology in content based video retrieval system. Although many researchers have been working on shot boundary detection for more than ten years, there are still a lot of problems to address.Works in this papers focus on shot boundary detection algorithm in MPEG compressed domain. We first review state-of-the-art algorithms on shot boundary detection(SBD), analyze their advantages and disadvantage. Then bring forward a new method to detect shot boundaries based on DC images and HSV color histogram in MPEG compressed domain. Based on first step, a qualitative analysis method of dominant motion about scheme compressed video stream based on the directional distribution of motion vectors and the feature of the relationship between any two motion vectors in MPEG2 stream is propose. It fits the motion vectors in MPEG stream into the two-dimensional affine model to detect basic camera operations automatically. We define a set of camera motion class that consists of six well-known basic camera operations:They are pan, tilt, object motion and static, respectively. The proposed approach is advantageous over the others in terms of the efficiency and robustness.