Research On Video Keyframe Extraction Technology Based On MPEG Compression Domain

Digital video has been applied in the form of compressed domain more and more comprehensive,and how to manage and search compressed video information effectively and quickly is one of the most important subjects of video retrival based on content. Keyframe extraction technology is the kernel of video retrival, so keyframe extraction based on MPEG compressed domain will be reseached in this paper.Firstly,at the beginning of this paper, the background and meaning of the research of this subject are introduced, and this paper gave the development history and current situation of video at abroad and home. The characteristics of digital video structure are introdueced, especially video structure in MPEG domain.Secondly,after we studied on the current key frame technologies stantively, compared these traditional methods, self-adaption clustering video keyframe frame extraction based on MPEG compressed domain method is put forward. The method mainly uses image characteristics information of Intraframe which is extracted using partly decoding in MPEG compressed domain video, and euclidean distance of the neighboring Intraframes, then juge the keyframe by self-adaption threshold value.Thirdly, as the application of the research result, the experiment platform of self-adaption clustering video keyframe extraction based on MPEG compressed domain by VC++6.0 is designed. In this system, MPEG video can be read and showed, certainly, the results of the test can be also showed.Lastly, Different types of video are used for testing,which contain news video,film video, cartoon video etc. Recall,precision and compression ratio are used to verify the result.The results of the experiments demonstrate that key frames, which are extracted by using the self-adaption clustering video keyframe extraction based on MPEG compressed domain, are more when there are more motion quantity and more cuts in the shot, and the key frames are less when there are less motion quantity and less cuts in the shot. The mehod do not need set the clustering center and iterations in advance, it is not only applicable to those videos which are appointed. The method is available for keyframe extraction in MPEG compressed domain.