We introduce a robust formulation of the shape correspondence problem under missing geometry and topological alterations with the language of functional maps. Using perturbation analysis, we show how removal of shape parts changes the Laplace-Beltrami eigenfunctions, and exploit it as a prior on the spectral representation of the correspondence. Our approach operates completely in the spectral domain, allowing to analyze shapes with constant complexity independent of their size, and yielding unprecedented results on challenging benchmarks.
This presentation is part of Minisymposium “MS16 - Topological Image Analysis: Methods, Algorithms, Applications (3 parts)”
organized by: Patrizio Frosini (University of Bologna) , Massimo Ferri (University of Bologna) , Claudia Landi (University of Modena and Reggio Emilia) .