Equi-affine Invariant Geometry for Shape Analysis

摘要: Traditional models of bendable surfaces are based on the exact or approximate invariance to deformations that do not tear stretch shape, leaving intact an intrinsic geometry associated with it. These geometries typically defined using either shortest path length (geodesic distance), properties heat diffusion (diffusion distance) surface. Both measures implicitly derived from metric induced by ambient Euclidean space. In this paper, we depart restrictive assumption observing a different choice results in richer set geometric invariants. We apply equi-affine for analyzing arbitrary shapes positive Gaussian curvature. The potential proposed framework is explored range applications such as shape matching and retrieval, symmetry detection, computation Voroni tessellation. show some analysis tasks, equi-affine-invariant often outperform their Euclidean-based counterparts. further explore facial anthropometry newborns. homogeneous group better captured metric.