Video Content Structure Technology To Achieve

The lack of structure in video impedes the development of new video applications. In this paper, we investigate video content structuring techniques systematically. Based on the research of the theory of video production and perception and combining the tasks of semantically annotation and syntactically segmentation, we make a thorough investigation of video content structuring techniques at low, mid- and high levels respectively. The main contribution is as following: We define the new research field of video content structuring by clarifying the basic problem, the appropriate approach and methodology of this domain, and also by putting forward a research framework which consists of syntactically segmentation and semantically annotation tasks at low, mid- and high levels respectively. We systematically investigate the general processes of video content producing and perception. We observe that the two processes actually observe the same basic rules, and we further discuss and summarize the indication of these rules towards video content structuring. We explore the low and mid-level video content structuring techniques comprehensively. In particular, we devise and improve some key algorithms including: video shot, audio shot, story unit, audio paragraph, flashlight, paragraph delimiter, advertisement, action story units detection and audio shot classification, etc. We investigate the high-level video content structuring techniques of news, sports, movies, and soap operas. In particular, we solve the problems of news story detection, translucent caption detection, role annotation and excitement annotation, etc. We develop and implement of VideoCAR workbench, which gives a sound support to the research of video content structuring techniques.