Video shot selection and content-based scene detection for automatic classification of TV sports news

Abstract

Visual retrieval systems need to store a large amount of digital video data. The new possibilities offered by the Internet as well as local networks have made video data publicly and relatively easy available, for example Internet video collections, TV shows archives, video-on-demand systems, personal video archives offered by many public Internet services, etc. Video data should be indexed, mainly using content-based indexing methods. Digital video is hierarchically structured. Video is composed of acts, sequences, scenes, shots, and finally of single frames. In the tests described in the paper a special software called the AVI – Automatic Video Indexer has been used to detect shots in tested videos. Then, the single, still frames from different time positions in the shots detected in ten TV sports news have been thoroughly examined for their usefulness in an automatic classification of TV sports news.