PicSOM - content-based image retrieval with self-organizing maps

We have developed a novel system for content-based image retrieval in large , unannotated databases. The system is called PicSOM, and it is based on tr ee structured self-organizing maps (TS-SOMs). Given a set of reference imag es, PicSOM is able to retrieve another set of images which are similar to t he given ones. Each TS-SOM is formed with a different image feature represe ntation like color, texture, or shape. A new technique introduced in PicSOM facilitates automatic combination of responses from multiple TS-SOMs and t heir hierarchical levels. This mechanism adapts to the user's preferences i n selecting which images resemble each other. Thus, the mechanism implement s a relevance feedback technique on content-based image retrieval. The imag e queries are performed through the World Wide Web and the queries are iter atively refined as the system exposes more images to the user. (C) 2000 Els evier Science B.V. All rights reserved.