Design of large intelligent image database systems

Image knowledge structure and an access method for retrieving images are two of the most important problems in the design of large intelligent image data base systems. In this article, we present a software architecture which utilizes an image knowledge structure, called 2D C+‐string, as the theoretical basis for intelligent image data base systems. Our architecture also provides an efficient access method to quickly locate the desired images by pruning a large percentage of nonpromising iconic indices represented by 2D C+‐strings. The 2D C+‐string spatial knowledge representation makes the image data base system become more intelligent, while the access method supports effective image retrieval from a large image data base without degrading the system's overall performance. Experimental results show that our access method is flexible enough to adapt to the change of image data base and very efficient in reducing the number of searches in similarity retrieval. © 1996 John Wiley & Sons, Inc.