Research On The Algorithm For Video Key Frame Extraction In Compressed Domain

With the rapid development and wide application of computer technology, multimedia technology and Internet, multimedia data has been applied more and more comprehensive, and the most complex multimedia data is video information. It becomes more and more difficult for people to manage and search video information efficiently and quickly because of the rapid growth of video information. Therefore, how to search and manage these video data effectively and quickly becomes a subject to be solved urgently. This paper will research the key frame extraction in video retrieval system emphatically.This paper studied mainly a video key frame extraction algorithm which based on compressed domain. It extracted directly the compressed video coding data and analyzed these data, and then it extracted the key frame of video data with fuzzy decision theory. According to this algorithm, an application system of the key frame extraction is designed and realized. The main contents of the paper are as follows:In this paper, firstly the background and meaning of this subject are given, and this paper introduced the development history and current situation of video retrieval at home and abroad. On this basis, the basic theory of video compression and MPEG video compression standards are studied seriously, and the hierarchical structure of video streams is analyzed in detail. In addition, the inter-frame different information which the macroblock layer contains in MPEG video stream is studied emphatically.Secondly, the key technology in CBVR(Content-Based Video Retrieval) system is studied, it contains lens detection, key frame extraction, scene extraction, video indexing and browsing. The key frame extraction is studied emphatically, and then on the basis of comparing the algorithms of Macroblock Differ, the algorithm improvement and innovation are done, and a key frame extraction method from MPEG video stream, which made use of macroblock-based motion prediction and compensation, is put forward.Then, the programming implementation which aimed to the above algorithm is given, and a content-based video key frame extraction system is completed, and experimental results show that the system's output is in line with the expected results. Finally, a summary of the work of the full text and my own understanding of the development of video retrieval is given, what's more, it pointing out the struggling direction for future work.