A Two-Stage Framework for Polygon Retrieval Using Minimum Circular Error Bound

Shape matching is an important issue in pattern recognition and it has many real world applications. We have proposed a two-stage framework for polygon retrieval 12, 11] which incorporates both qualitative and quantitative measures of polygons in the rst and second stage respectively. In this paper, we introduce an extension to our two-stage framework. We propose a new polygon matching technique using Circle Error Bound and describe how this technique works under translation and scaling of polygons. Base on this technique, we propose a new translation invariant similarity measure for polygons named Minimum Circular Error Bound, which can be used in the second stage of the two-stage framework. From observation, the Minimum Circular Error Bound method produces similarity ranking of polygons resembling human perception. We compare the Minimum Circular Error Bound method with the Hausdorr Distance method and demonstrate the advantages of our method.