A Novel Shape Descriptor Based on Extreme Curvature Scale Space Map Approach for Efficient Shape Similarity Retrieval

The main drawbacks of Curvature Scale Space (CSS) matching are due to the problem of shallow and deep concavities on the shape. To solve this problem, in this paper we present a novel shape descriptor based on Extreme Curvature Scale Space (ECSS) map approach. Unlike the CSS map of shape which results from zeros crossings values of the curvature, the ECSS map is created by tracking the position of extreme curvature points. Similarly to CSS descriptor, our proposed one is based on the maxima of the obtained ECSS map. It is robust with respect to noise, scale and orientation changes of the shape. Several experiments have been conducted on a SQUID database. The obtained results prove the efficiency of the proposed shape descriptor when is compared to the CSS one, especially in the case of shallow or deep concavities.