VIDSEEK: Dynamic Multi-dimensional Browsing of Video Archives

We describe a dynamic Web-based digital video library browsing system (VIDSEEK) that allows users to preview the contents of the VISION digital video library via automatically selected and organized keyframes. Our system supports dynamic clustering-on-demand, which allows users organize the digital video library clips into clusters based on multiple user-specified video features. The system also supports category-based browsing, which allows users to interactively and dynamically filter the VISION digital video library clips based on a given set of constraints, such as video source, keywords, and date of capture. We use efficient techniques to create the video abstractions, and dynamic organization, i.e., categorization or clustering, of the video abstractions to provide a sophisticated tool for video archive exploration.