Non-parametric planar shape representation based on adaptive curvature functions

This paper presents a non-parametric method to extract a very short feature vector from the curvature function of a planar shape. Curvature is adaptiv ely calculated using a new procedure that removes noise from the contour wi thout missing relevant points. Then, its Fourier transform is projected ont o a set of vectors, which have been chosen to be as representative as possi ble, to obtain the similarity between the input object and each vector of t he set. These similarity values are the elements of the feature vector. The proposed method is very fast and classification has proven that the repres entation is good. (C) 2001 Pattern Recognition Society. Published by Elsevi er Science Ltd. All rights reserved.