Sub-sampling for efficient spectral mesh processing

摘要: In this paper, we apply Nystrom method, a sub-sampling and reconstruction technique, to speed up spectral mesh processing. We first relate method Kernel Principal Component Analysis (KPCA). This enables us derive novel measure in the form of matrix trace, based soly on sampled data, quantify quality approximation. The is efficient compute, well-grounded context KPCA, leads directly greedy sampling scheme via trace maximization. On other hand, analyses show that it also motivates use max-min farthest point sampling, which more alternative. demonstrate effectiveness with compared random using two applications: segmentation correspondence.