A content based shape retrieval system

This paper outlines a methodology of building a content based image storage and retrieval system by extracting and querying shape related information from the image and representing them as vectors in a database. The method is based on the popular centroid-radii model, however to make it invariant to rotation this paper proposes comparing the number of radii of specific lengths instead of directly comparing the lengths themselves. The normalization of radii lengths also makes the process invariant to scaling. To represent shapes with holes inside them the method is extended by considering a number of concentric circles centered at the centroid and computing radii histograms for the portion of the shape within each of the circles. The similarity of two shapes are judged by comparing the number of radii in corresponding ranges. The method has been seen to yield an appreciable amount of accuracy