Mining large-scale broadcast video archives towards inter-video structuring
PCM'04 Proceedings of the 5th Pacific Rim Conference on Advances in Multimedia Information Processing - Volume Part II
| Hi-index | 0.00 |
Based on a simple temporal structural model of news program, this paper presents a practical solution to automatic news story segmentation by integrating syntactic and semantic methods. First, a syntactic segmentation method is used to detect the shot boundaries in order to partition video frames into video shots. Then a semantic segmentation method based on the graph-theoretical cluster analysis is developed to classify the video shots into anchorpersonshots and news footage shots. Finally, a structural model of news video is used to complete the news-story segmentation. The proposed method obtains a precision of 90.45% and a recall of 95.83%in the segmentation experiment of 168 news stories from two Hong Kong news stations.