A hypergraph-based image database clustering framework

This paper describes a new approach to image database clustering. The method requires no a priori information. It works free of context and previous knowledge: in a first stage, the image features are formed automatically, and modeled by a p-Nearest Neighbor Hypergraph (p-NNH) representation. Then images are clustered to form categories using a multilevel p-NNH partitioning approach. The partitioning approach operates on Coarsening-Paritioning-UnCoarsening scheme (CPUC). Categories are visualized by displaying the most typical image(s) of the categories as thumbnails. The main benefit of the approach is that it deals with a large volume image database and with a representation structure (hypergraph) that is close to the human visual grouping system. To judge results, an evaluation scheme which is adequate for the task of categorization is proposed.