Automatic detection of headlines in temporally aggregated TV sports news videos

Abstract

Internet video collections, TV shows archives, video-on-demand systems, video libraries, personal video archives offered in the Web, etc. store more and more amount of digital video data. Therefore, new methods for automatic indexing and retrieval of video data are being developed, new technologies are being proposed. Due to a huge commercial appeal the TV sports news has become one of the dominant application areas for video automatic indexing and retrieval. Content-based indexing of TV sports news usually starts by the automatic segmentation, then recognition and classification of scenes reporting sports events. The detection of video structure, not only of the simplest structural elements like frames or shots but also of video scenes, leads to the optimization of indexing process. The automatic identification of sports disciplines in TV sports news will be less time consuming if the analyzed video material is limited to player scenes included in news headlines. The temporal aggregation process facilitates the detection of headline shots. The paper presents a method of automatic detection of sports news headlines for content-based indexing of TV sports news based on video structure information.