Content-based image retrieval using composite features

In this paper we demonstrate methods for content-based image retrieval problems utilizing a specially designed retrieval architecture implemented within the 'Intelligent Image Retrieval' system (IIR). The method consists Of new image features in the retrieval context, including segmentation methods and use of image frames. They are combined in a unique way with color, texture, shape and localization information with a special data abstraction construction in a graphical user interface. The IIR System provides an efficient retrieval architecture utilizing a tailored query language, retrieval mechanisms and an object-oriented database enabling the use of complex data structures and relations needed for successful query processing Functionality and performance of methods and architecture are illustrated with a series of tests using a database that consists of several hundred 'scene' images.