Intrinsic Mesh Segmentation

Mesh segmentation offers a desirable divide-and-conquer strategy for many graphics applications. In this paper, we present a novel, efficient, and intrinsic method to segment meshes following the minima rule. The eigenfunctions of the Laplace-Beltrami operator define locality and volume-shape preserving functions over a surface. Inspired on Manifold learning theory, we use these functions as the basis of an embedding space for mesh vertices and group them using k-means clustering. We also present a new kind of segmentation hierarchy built from the analysis of the Laplace-Beltrami operator spectrum.