Technology Research, Content-based Video Retrieval

Application of the video information is more and more extensive,such as shopping online,video webstation,VOD and digital library,etc. The technology among them is mainly concentrated on these respects:video compressing,transmitting,management,control and searching,etc. Because video information is heavy,and abundant in content,video effective management and search becomes a difficult subject.Today,the management and search of video is done mainly based on video file and its describing,which can't operate from the intra content of video ,however,it is usally what the user cares. So occurse a new field about video processing:content-based analyse of video,that can be called "video understanding".This paper have done the following groundwork around this field:1 Because video data is non-structed data,there is not a clear structure in video data,that makes the video analysis difficult. So get a rational structure model is very necessary. In this paper,followed analysed the characteristic of the existing video model,a comparatively suitable video model has been put forward according to MPEG7 standard. All of the other work later are launched on this model foundation.2 Shot change detection is a foundation work in content based video analysis. In this paper,the most algorithms are studied,and pointed out the merit and shortcoming,some improvement is done. It give a algorithm based on twin-comparison and sample-frame technology. Another algorithm is put forward that be adaptability to frame excursion and luminance strengthen that has not considered in most algorithms .3 video rearch is a basic demand for user,which belongs to later stage .This task needs analysing in earlier stage such as shot detection and character extraction which are not solved ideally at present,and this task is studied lesser. In this paper,Some existing methods are analysed,and various kinds of search modes are discussed. A way of extracting keyframe called local "shot" dectetion is put forward which can present the shot in a idea way. At last have proposed a suitable video clip searching algorithm for television advertising.4 It designs a system including shot detection and video clip retrieval,supported by the algorithm above. It is really a proof of the refered technology.