Video Content Browsing Based on Iterative Feature Clustering for Rushes Exploitation
Werner Bailer, Christian Schober, Georg Thallinger · TRECVID · 2006
Rushes Exploitation We have implemented a feature-independent framework for video content browsing, which is based on iterative clustering and filtering of the content set. Plug-ins for content clustering using the features camera motion, motion activity, audio volume, face occurrences, global color similarity and object similarity have been implemented. A light table view is used for visualization in the browsing tool. We have evaluated the tool on the TRECVID rushes data with four browsing tasks defined by a textual description of the material. We measured precision, recall and the number of items found during the working time.