Color and spatial feature for content-based image retrieval

Most of the currently available image database systems provide a text-based retrieval function called keyword retrieval, where users specify 'keywords ' such as titles, attributes, and categories of themes. But many times it i s not easy for users to specify suitable keywords for a particular retrieva l. Besides, building a large image database with complete description of co ntents is a very difficult task. In this paper, we present a content-based retrieval method which obviates the need to describe certain contents of an image to be archived and retrieved. The proposed method computes image fea tures automatically from a given image and they can be used to archive and/ or retrieve images. These features are based on color and its spatial distr ibution information in an image. We have also developed a similarity measur e to compare the color and spatial feature similarity of two images. This t echnique has been developed and tested for content-based similarity retriev al of images on two databases consisting of: (i) 100 test images and (ii) 8 00 actual trademarks images. The experimental results show a high efficienc y of retrieval. (C) 1999 Published by Elsevier Science B.V, All rights rese rved.