Shape analysis using the auto diffusion function

摘要: Scalar functions defined on manifold triangle meshes is a starting point for many geometry processing algorithms such as mesh parametrization, skeletonization, and segmentation. In this paper, we propose the Auto Diffusion Function (ADF) which linear combination of eigenfunctions Laplace-Beltrami operator in way that has simple physical interpretation. The ADF given 3D object number further desirable properties: Its extrema are generally at tips features object, its gradients level sets follow or encircle features, respectively, it controlled by single parameter can be interpreted feature scale, and, finally, invariant to rigid isometric deformations. We describe properties detail compare other choices scalar manifolds. As an example application, present pose invariant, hierarchical skeletonization segmentation algorithm makes direct use ADF.