A region-based image representation for spatial reasoning and similarity retrieval in image database systems

In this paper we propose a region-based shape representation image database. We extend Guojun's method for representing an object to a representation of any picture that contains one or more objects. This system contains some functions like spatial reasoning, visualization, browsing and similarity retrieval and so on. A new vector-based method for spatial reasoning is also proposed in this paper. In addition to the advantages of good precision of Guojun's model, which required lower storage requirement than the FD-based method, our method has other advantages, for example, the abilities to represent any picture that contains more than one object, to reconstruct pictures and spatial reasoning.