Concept-based browsing in video libraries

Digital library applications based on huge amounts of digital video data require efficient browsing and searching mechanisms for the extraction of relevant information. To avoid information overload, a browsing system needs to preselect shots of interest from the database in a user-adequate manner. A retrieval engine for video browsing is proposed that offers conceptual, content-based access to videos. It calculates relevance values for the results of a conceptual query by feature aggregation on video shot granularity. This engine is embedded in a browsing system architecture which was extended with an intelligent client buffer strategy and admission control mechanism aiming for browsing specific requirements. Thus, we support continuous presentation of time-dependent media and reduce startup latency.