Curvature estimation of point-sampled surfaces and its applications

Abstract

In this paper, we propose a new approach to estimate curvature information of point-sampled surfaces. We estimate curvatures in terms of the extremal points of a one-dimensional energy function for discrete surfels (points equipped with normals) and a multi-dimensional energy function for discrete unstructured point clouds. Experimental results indicate that our approaches can estimate curvatures faithfully, and reflect the subtle curvature variations. Some applications for curvature information, such as surface simplification and feature extraction for point-sampled surfaces, are given. © Springer-Verlag Berlin Heidelberg 2005.