Spectral Distance Distributions for Non-rigid Objects

摘要

Non-rigid shape deformation without tearing or stretching is called isometry. There are many difficulties to research non-rigid shape in Euclidean space. Therefore, non-rigid shapes are firstly embedded into a none-Euclidean space. Spectral space is chosen in this paper. Then three descriptors are proposed based on three spectral distances. The existence of zero-eigenvalue has negative effects on computation of spectral distance. Therefore the spectral distance should be computed from the first non-zero-eigenvalue. Experiments show that spectral distance distributions are very effective to describe the non-rigid shapes.