Fuzzy Spatial Position between Objects in Image Retrieval

As we all know, ambiguity of spatial position between objects is still a difficult problem for image classification and retrieval. In this paper, a fuzzy spatial position definition method between objects based on Markov Random Field (MRF) is presented. This framework is a kind of fuzzy pattern-matching approach, which consists of comparing an object to a fuzzy set and representing the degree of position satisfaction to a reference object, moreover, the intricate relationships can be represented by Fuzzy Attributed Relational Graphs (FARGs) in this method. Finally, by comparing the performance of this approach with conventional techniques, effectiveness of this method to content-based image retrieval is illuminated.