English

Implicit Manifold Reconstruction

Computational Geometry 2019-04-09 v1

Abstract

Let MRd{\cal M} \subset \mathbb{R}^d be a compact, smooth and boundaryless manifold with dimension mm and unit reach. We show how to construct a function φ:RdRdm\varphi: \mathbb{R}^d \rightarrow \mathbb{R}^{d-m} from a uniform (ε,κ)(\varepsilon,\kappa)-sample PP of M\cal M that offers several guarantees. Let ZφZ_\varphi denote the zero set of φ\varphi. Let M^\widehat{{\cal M}} denote the set of points at distance ε\varepsilon or less from M\cal M. There exists ε0(0,1)\varepsilon_0 \in (0,1) that decreases as dd increases such that if εε0\varepsilon \leq \varepsilon_0, the following guarantees hold. First, ZφM^Z_\varphi \cap \widehat{\cal M} is a faithful approximation of M\cal M in the sense that ZφM^Z_\varphi \cap \widehat{\cal M} is homeomorphic to M\cal M, the Hausdorff distance between ZφM^Z_\varphi \cap \widehat{\cal M} and M\cal M is O(m5/2ε2)O(m^{5/2}\varepsilon^{2}), and the normal spaces at nearby points in ZφM^Z_\varphi \cap \widehat{\cal M} and M\cal M make an angle O(m2κε)O(m^2\sqrt{\kappa\varepsilon}). Second, φ\varphi has local support; in particular, the value of φ\varphi at a point is affected only by sample points in PP that lie within a distance of O(mε)O(m\varepsilon). Third, we give a projection operator that only uses sample points in PP at distance O(mε)O(m\varepsilon) from the initial point. The projection operator maps any initial point near PP onto ZφM^Z_\varphi \cap \widehat{\cal M} in the limit by repeated applications.

Cite

@article{arxiv.1904.03764,
  title  = {Implicit Manifold Reconstruction},
  author = {Siu-Wing Cheng and Man-Kwun Chiu},
  journal= {arXiv preprint arXiv:1904.03764},
  year   = {2019}
}

Comments

A preliminary version appears in Proceedings of the 25th Annual ACM-SIAM Symposium on Discrete Algorithms, 2014, 161--173