Database Support for Multimedia Retrieval

More and more multimedia applications require effective and efficient support for similarity search in large databases, i.e. fast retrieval of objects that are similar to a given query object with respect to particular features including shape, appearance, color distribution, or scene composition to name a few. After a short survey of 3D shape similarity search based on shape histograms along with algorithms for efficient processing of ellipsoid queries, we discuss the topic of relational indexing. In order to obtain industrial strength, query processing has to be integrated into existing robust database systems. The relational interval tree represents an example for relationally managing interval data, and we plan to adopt this idea to similarity search. The presentation is concluded by sketching future research plans including new usability models for similarity search as well as addressing new complex distance functions.

Authors: Seidl T.
Published in: In: Klein R., Clausen M. (eds.): Perspective Seminar Multimedia Retrieval. Dagstuhl Seminar 03112
Sprache: EN
Jahr: 2003
Typ: Sonstige
Forschungsgebiet: Exploration of Multimedia Databases