Self-organizing Maps for Content-based Image Database Retrieval

We have developed a novel system for retrieving images similar to a given set of reference images in large image databases, based on Tree Structured Self-Organizing Maps (TS-SOMs). Our image retrieval system is called PicSOM. It has been designed with the purpose to provide a framework for generic research on algorithms and methods for content-based image retrieval. A new technique introduced in this paper facilitates automatic combination of the responses from multiple TS-SOMs and their hierarchical levels. Each TS-SOM is tuned with a diierent image feature representation like color, texture, or shape. This mechanism adapts to the user's preferences in selecting which images resemble each other, in the particular sense the user is interested in. The image queries are performed through the World Wide Web and the queries are iteratively reened as the system exposes more images to the user.