Research On Key Frame Extraction Algorithm Of MPEG Video In Compressed Domain Based On Self-adaptive Threshold

In recent years,the rapid development of computer technology and multimedia technology have brought greate changes to people's life,people receive a large amount of information every day,while video data occupy a large proportion in multimedia information.With the increase of video data,convenient to search video clips of interest have become increasingly urgent to people, therefore, video database management system gradually become the research hotspots.The video retrieval base on the content has currently become one of the key technologies of a video data management information system. While the key frame extraction technology is the foundation of video analysis and video retrieval base on the content. The use of key frame not noly significantly reduces the data of video retrieval, but also provides an organizational framework for video retrieval and summary. The efficiency of key frame extraction methods and video retrieval directly relates to the efficiency of video retrieval system, thus research of key frame extraction methods is of great significanceThere are two types of methods to extract key frame, one is to extract key frame base on uncompressed domain in which each frame will need to be decoded completely, and the other is to extract them base on compressed domain in which each frame will only need to be decoded partly. This thesis (design) firstly analyzes the current situation of key frame extraction methods, and then deeply researches common methods of key frame extraction, as well as the international standards of MPEG. Finally, author programs and realized an algorithm of extracting key frame of MPEG video in compressed domain based on self-adaptive threshold. The algorithm firstly filters B frames and P frames, only using I frames for images' features extraction, which simplifies the process of extracting key frames, and then segment shot as well as extract images'features, which greatly reduce the computational complexity and improve the efficiency of the extraction of key frames. Moreover, this algorithm using self-adapting threshold, which better adapt to all types of MPEG compressed movies, overcome defects of common algorithm that process only a class of film.