G. Manson and S.-A. Berrani (France)
Multimedia, Video Technology, TV Broadcast structuring,Indexing, TV-on-Demand.
This paper addresses the problem of TV stream macro-segmentation which consists of automatically determining the start and the end of each broadcasted program and inter program (e.g. commercial, trailer). As programs do not share any common features, this paper focuses in particular on detecting inter-programs. Programs are then extracted as the rest of the stream. More precisely, inter programs are detected by their repetition property. This paper shows in a experimental way that almost all inter programs are broadcasted several times in the stream and they are repeated sufficiently in order to achieve an accurate macro-segmentation. It also presents a repeated sequence technique and a rule-based technique that performs macro-segmentation using detected repeated sequences. Our macro-segmentation solution is completely automatic, it outperforms the metadata macro-segmentation that comes with the TV stream and it requires less than two days of accumulated TV stream. The effectiveness of the approach is shown experimentally on a real continuous 7 days TV broadcast.
Important Links:
Go Back
