An Efficient Content-Based Retrieval System for Large Image Database

Abstract

Content-based image retrieval has become more desirable for developing large image databases. This chapter presents an efficient method of retrieving images from an image database. This system combines color, shape and spatial features to index and measure the similarity of images. Several color spaces that are widely used in computer graphics are discussed and compared for color clustering. In addition, this chapter proposes a new automatic indexing scheme of image databases according to our clustering method and color sensation, which could be used to retrieve images efficiently. As a technical contribution, a Seed-Filling like algorithm that could extract the shape and spatial relationship feature of an image is proposed. Due to the difficulty of determining how far objects are separated, this system uses qualitative spatial relations to analyze object similarity. Also, the system is incorporated with a visual interface and a set of tools, which allows the users to express the query by specifying or sketching the images conveniently. The feedback learning mechanism enhances the precision of retrieval. The experience shows that the system is able to retrieve image information efficiently by the proposed approaches.