Enhancing CSS-based shape retrieval for objects with shallow concavities

Curvature scale space (CSS) image is a multi-scale organisation of the infl ection points of a closed planar curve as it is smoothed. It consists of se veral arch shape contours, each related to a concavity or a convexity of th e curve. The maxima of these contours have already been used as shape descr iptors to find similar shapes in large image databases. In this article, we address the problem of shallow concavities. These may give rise to large c ontours in the CSS image. These contours may then match those corresponding to deep and wide concavities during the matching process. The phenomenon c an be explained by recalling the fact that Gaussian smoothing leads to an a pproximation of geometric heat equation deformation. We have introduced a m ethod to enrich the CSS image and create different contours for different t ypes of concavities. We tested the proposed method on a database of 1100 im ages of marine creatures. A significant improvement was observed in the per formance of the system on shapes with shallow segments. (C) 2000 Elsevier S cience B.V. All rights reserved.