K-Mutual Nearest Neighbour Approach for Clustering Two-Dimensional Shapes Described by Fuzzy-Symbolic Features

In this paper, a new method of representing two-dimensional shapes using fuzzy-symbolic features and a similarity measure defined over fuzzy-symbolic features useful for clustering shapes is proposed. A k-mutual nearest neighborhood approach for clustering two-dimensional shapes is presented. The proposed shape representation scheme is invariant to similarity transformations and the clustering method exploits the mutual closeness among shapes for clustering. The feasibility of the proposed methodology is demonstrated by conducting experiments on a considerably large database of shapes and also, its validity is tested by comparing with the well known clustering methodologies.