Retrieval of Pictures Using Approximate Matching

this paper, we describe a general-purpose pictorial retrieval system based on approximate matching. This system accommodates pictorial databases for a broad class of applications. It consists of tools for handling the following aspects--- user interfaces, reasoning about spatial relationships, computing degrees of similarity between queries and pictures. In this paper, we briefly describe the model that is used for representing pictures/queries, the user interface, the system for reasoning about spatial relationships, and the methods employed for computation of similarities of pictures with respect to queries. We assume that there is a database containing the pictures. We also assume that there is some meta-data associated with each picture which describes the contents of the picture. This meta-data contains information about the objects in the picture, their properties and the relationships among them. For example, consider a picture containing a man shaking hands with a woman and is to the left of the woman. The meta-data about this picture identifies two objects, man and woman, and the spatial relationshi