The paper introduces a novel approach for interactive video browsing that makes video content fully transparent to the user. Video clips are analysed and indexed by two tree structures: a content index tree representing the content of automatically segmented video shots and a time index tree representing the temporal structure. The index top levels give an overview over the entire content. Subsequent levels illustrate content relationships more detailed. Every level of both trees is a two-dimensional self-organising map organising media objects by two degrees of freedom. Media objects are represented by content-based visual MPEG-7 descriptions. The implemented navigation scheme allows the user for switching between content index tree and time index tree without loosing the overview. Context information (position in the tree, content of next lower level, etc.) is permanently shown in auxiliary panels. The implementation is based on the scalable vector graphics standard (visualisation) and the MPEG-7 reference implementation. First evaluation results show that the proposed approach facilitates accessing video content in a novel way.
H. Eidenberger: "A Video Browsing Application based on visual MPEG-7 Descriptors and Self-Organising Maps"; International Journal of Fuzzy Systems, 3 (2004).
Click into the text area and press Ctrl+A/Ctrl+C or ⌘+A/⌘+C to copy the BibTeX into your clipboard… or download the BibTeX.