Decentralized Image Retrieval for Education (DIRECT)

University Of Virginia Main Campus, Charlottesville VA

Investigators

Abstract

The Decentralized Image Retrieval for Education (DIRECT) project is developing a peer-to-peer content based image retrieval (CBIR) service for the National SMETE Digital Library program. CBIR allows the user to designate a query image so that the service can search the library for images of similar content. DIRECT matches images not by text metadata but by the color, texture, and shape of the image objects. With such a system the users of the NSDL do not need to know specialized languages to initiate a search. Furthermore, the matching process does not depend on a match between the cataloguer description and the user description. The system is available to all collections in the NSDL without imposing new standards or protocols. This offers the promise for the NSDL to support images that have not yet been cataloged or have incomplete metadata, without the image collection provider or aggregator having to incur additional cataloging overhead.