Shot classification in sports video

Shot classification and transition context analysis are fundamental work for video structure and content analysis. In this paper, we present an effective hierarchical shot classification scheme based on replay and nonreplay as viewed from video editing and semantics in sports video. The scheme facilitates video semantics analysis with giving prominence to replay. Firstly, a shot is categorized into replay or nonreplay class by replay-logo pairing. The replay-logo can be automatically extracted from the sports video in advance. Then, nonreplay shots are further classified into long, medium, close-up or out-field type, each one of which indicates certain semantics respectively. Every type of shot is intrinsically characterized by effective features. Extensive experiments have demonstrated the method's validity and effectiveness.