摘要: Spectral compression of the geometry triangle meshes achieves good results in practice, but there has been little or no theoretical support for optimality this compression. We show that, certain classes geometric mesh models, spectral decomposition using eigenvectors symmetric Laplacian connectivity graph is equivalent to principal component analysis on that class, when equipped with a natural probability distribution. Our proof treats connected one-and two-dimensional fixed convex boundaries, and based an asymptotic approximation distribution case. The key identical, up constant factor, inverse covariance matrix valid geometries. Hence, optimal, mean square error sense, these under some assumptions their