A survey of access methods for image data

Abstract

The feasibility of storing large amount of digital images and the pictorial-oriented nature of many applications have attracted much interest in pictorial information systems. An important issue in such systems is the efficient retrieval of image based on its contents. Supporting content-based retrieval of image data is a difficult problem and embraces different technologies including image processing, user interface design, and database management. To provide efficient content-based retrieval, efficient access methods based on image features are required. This paper presents a survey of indexing techniques that support content-based retrieval of images. In particular, we examine access methods for three types of image features, namely, object shape, semantic objects, and spatial relationships among semantic objects. For each image feature, we look at how various approaches represent and organize the image features and how they support similarity retrieval.