Search

Typically, multimedia database management systems process content-based image retrieval queries by extracting a set of features from each data object as it is inserted into the underlying database. By expressing queries that are based upon these features, users are able to retrieve the data objects back from the database. Previous research has demonstrated that one method of improving the effectiveness of similarity searches in such systems is to augment the underlying database with a set of edited images to allow more flexible matching. Space can be saved by storing the additional images as sequences of editing operations instead of as large binary objects. This paper proposes an approach for processing retrieval queries in such an environment and presents the results of a performance evaluation demonstrating the effectiveness of the approach.