Browsing Sports Video

he case for automated video information access systems has been made repeatedly by numerousresearchers since the early 1990s. Finding important episodes in nontext media like video andaudio is made difficult because the same material means different things to different people.Therefore, manually tagging data with textual keywords as a means of facilitating access is notnecessarily reliable for this type of media. This is a crucial point. Information retrieval as a task isreally only defined in the context of the user. This article considers one such context: sports media. In thatcontext, there is a salient structure to the media as well as a series of definable user requirements. Sportbroadcasting is certainly of high commercial importance [1]. All national and international news broadcastscontain specific regular segments devoted to sports. Consumers themselves are increasingly acting as genera-tors of sport media as it becomes easier for parents and coaches to record school sports games, for instance. From a broadcaster’s point of view, efficient archiving of media facilitates later reuse in the creation ofhighlights packages or specialized DVDs. Manual keywording of events is feasible but time consuming and