No abstract

No abstract

No abstract

Content based video organization requires to understand semantical relationships such as identity and similarity between video segments. In particular, it is of great interest to identify scenes of the same event which may differ in their appearances. For broadcasted sports video, such scenes correspond to live and replay scenes that appear at different temporal positions. In this paper, we propose a method of linking up live and replay scenes by focusing on the domain knowledge about producing TV programs of sports: most of the replay scenes are sandwiched in between specific digital video effects. The replay scenes are linked with live scenes when the game is in play based on salient image features. We clarify the effectiveness of our method through fundamental experiments for American football games.