Abstract
Given a set of points that sample a shape, the Rips complex of the points is often used to provide an approximation of the shape easily-computed. It has been proved that the Rips complex captures the homotopy type of the shape, assuming that the vertices of the complex meet some mild sampling conditions. Unfortunately, the Rips complex is generally high-dimensional. To remedy this problem, it is tempting to simplify it through a sequence of collapses. Ideally, we would like to end up with a triangulation of the shape. Experiments suggest that, as we simplify the complex by iteratively collapsing faces, it should indeed be possible to avoid entering a dead end such as the famous Bing’s house with two rooms. This paper provides a theoretical justification for this empirical observation. We demonstrate that the Rips complex of a point cloud (for a well-chosen scale parameter) can always be turned into a simplicial complex homeomorphic to the shape by a sequence of collapses, assuming that the shape is nicely triangulable and well-sampled (two concepts we will explain in the paper). To establish our result, we rely on a recent work which gives conditions under which the Rips complex can be converted into a Čech complex by a sequence of collapses. We proceed in two phases. Starting from the Čech complex, we first produce a sequence of collapses that arrives to the Čech complex, restricted by the shape. We then apply a sequence of collapses that transforms the result into the nerve of some covering of the shape. Along the way, we establish results which are of independent interest. First, we show that the reach of a shape cannot decrease when intersected with a (possibly infinite) collection of balls, assuming the balls are small enough. Under the same hypotheses, we show that the restriction of a shape with respect to an intersection of balls is either empty or contractible. We also provide conditions under which the nerve of a family of compact sets undergoes collapses as the compact sets evolve over time. We believe conditions are general enough to be useful in other contexts as well.
We’re sorry, something doesn't seem to be working properly.
Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.
References
Adiprasito, K., Benedetti, B.: Metric geometry and collapsibility. arXiv:1107.5789 (2011)
Amenta, N., Bern, M.: Surface reconstruction by Voronoi filtering. Discrete Comput. Geom. 22(4), 481–504 (1999)
Attali, D., Lieutier, A.: Reconstructing shapes with guarantees by unions of convex sets. In: Proceedings of the 26th Annual Symposium on Computational Geometry, Snowbird, pp. 344–353 (2010)
Attali, D., Edelsbrunner, H., Mileyko, Y.: Weak witnesses for Delaunay triangulations of submanifold. In: ACM Symposium on Solid and Physical Modeling, pp. 143–150. Beijing (2007)
Attali, D., Lieutier, A., Salinas, D.: Efficient data structure for representing and simplifying simplicial complexes in high dimensions. Int. J. Comput. Geom. Appl. 22(4), 279–303 (2012)
Attali, D., Lieutier, A., Salinas, D.: Vietoris–Rips complexes also provide topologically correct reconstructions of sampled shapes. Comput. Geom. 46(4), 448–465 (2013)
Björner, A.: Topological methods. In: Handbook of Combinatorics, vol. II, pp. 1819–1872. North-Holland, Amsterdam (1995)
Boissonnat, J.-D., Cazals, F.: Natural neighbor coordinates of points on a surface. Comput. Geom. 19(2), 155–173 (2001)
Carlsson, E., Carlsson, G., De Silva, V.: An algebraic topological method for feature identification. Int. J. Comput. Geom. Appl. 16(4), 291–314 (2006)
Chambers, E., De Silva, V., Erickson, J., Ghrist, R.: Vietoris–Rips complexes of planar point sets. Discrete Comput. Geom. 44(1), 75–90 (2010)
Chazal, F., Lieutier, A.: Smooth manifold reconstruction from noisy and non uniform approximation with guarantees. Comput. Geom. 40, 156–170 (2008)
Chazal, F., Cohen-Steiner, D., Lieutier, A.: A sampling theory for compact sets in Euclidean space. Discrete Comput. Geom. 41(3), 461–479 (2009)
Cohen, M.M.: A Course in Simple-Homotopy Theory. Springer-Verlag, New York (1973)
de Silva, V., Ghrist, R.: Coverage in sensor networks via persistent homology. Algebr. Geom. Topol. 7, 339–358 (2007)
Delfour, M.C., Zolésio, J.-P.: Shapes and Geometries: Metrics, Analysis, Differential Calculus, and Optimization. SIAM, Philadelphia (2011)
Federer, H.: Curvature measures. Trans. Am. Math. Soc. 93, 418–491 (1959)
Hausmann, J.: On the Vietoris–Rips complexes and a cohomology theory for metric spaces. Ann. Math. Stud. 138, 175–188 (1995)
Latschev, J.: Vietoris–Rips complexes of metric spaces near a closed Riemannian manifold. Arch. Math. 77(6), 522–528 (2001)
Munkres, J.: Elements of Algebraic Topology. Perseus Books, New York (1993)
Niyogi, P., Smale, S., Weinberger, S.: Finding the homology of submanifolds with high confidence from random samples. Discrete Comput. Geom. 39(1–3), 419–441 (2008)
Tahbaz-Salehi, A., Jadbabaie, A.: Distributed coverage verification in sensor networks without location information. IEEE Trans. Autom. Control 55(8), 1837–1849 (2010)
Whitney, H.: Geometric Integration Theory. Dover Publications, New York (2005)
Acknowledgments
We thank the anonymous reviewers for many helpful suggestions. Research partially supported by the French “Agence nationale pour la Recherche” under Grant ANR-13-BS01-0008 TopData.
Author information
Authors and Affiliations
Corresponding author
Additional information
Editor in Charge: Herbert Edelsbrunner
Appendix: \(C^{1,1}\) diffeomorphisms preserve nicely triangulable manifolds
Appendix: \(C^{1,1}\) diffeomorphisms preserve nicely triangulable manifolds
The goal of this section is to prove Theorem 6.
Proof of Theorem 6
Let \(x \in M\). Since M is a compact \(C^{1,1} k\)-manifold embedded in \(\mathbb {R}^d\), there exists a k-dimensional affine space \(T_M(x) \subset \mathbb {R}^d\) tangent to M at x. Let \(\pi _x : M \rightarrow T_M(x)\) be the orthogonal projection onto the tangent space \(T_M(x)\) and let \(\pi _{\Phi (x)}' : M' \rightarrow T_{M'}(\Phi (x))\) the orthogonal projection onto \(T_{M'}(\Phi (x))\). Since M and \(M'\) are compact, we can find two constants K and \(K'\) independent of x such that:
Given \(t_0 > 0\), we consider the open set \(U_x = M \cap {\text {B}}^\circ (x, t_0)\) and adjust \(t_0\) in such a way that
-
1.
The restriction \(\pi _x : U_x \rightarrow \pi _x(U_x)\) is a homeomorphism for all \(x \in M;\)
-
2.
The restriction \(\pi _{\Phi (x)}' : \Phi (U_x) \rightarrow \pi _{\Phi (x)}'(\Phi (U_x))\) is also a homeomorphism for all \(x \in M\).
For sake of conciseness, we only sketch a justification for the existence of such a \(t_0>0\). The local property (i.e. the existence of \(t_0>0\) for a given \(x \in M\)) follows easily from the definition of embedded \(C^{1,1} k\)-manifolds. Indeed, the assumption of a regular embedding entails that \(\pi _x\) has full rank derivative at x and the inverse function theorem can be applied to get the local property. In order to get a uniform \(t_0>0\) (the requested global property) one can establish first the following strengthening of the local property: For any \(x \in M\), there is \(t_x>0\) such that for any \(y \in M \cap B^\circ (x, t_x)\), the restriction of \(\pi _y\) to \(M \cap B^\circ (x, t_x)\) is a \(C^1\) homeomorphism. Compactness of M can then be used in the usual manner to get a uniform \(t_0\).
The collection of pairs \(\{(U_x,\pi _x)\}_{x \in M}\) forms an atlas in the \(C^{1,1}\) structure of M. Similarly, the collection of pairs \(\{(\Phi (U_x),\pi _{\Phi (x)}')\}_{x \in M}\) forms an atlas in the \(C^{1,1}\) structure of \(M'\). Let \({\mathcal {T}}_M(x)\) be the linear space associated to \(T_M(x)\) and denote by \(D\Phi _x\) the derivative of \(\Phi \) at x, seen as a linear map between \({\mathcal {T}}_M(x)\) and \({\mathcal {T}}_{M'}(\Phi (x))\). Since M is compact, there is a constant \(K_\Phi \) independent of x such that
and two positive numbers \(\kappa _2 \ge \kappa _1 > 0\), again independent of x by compactness of M, such that
Consider the affine function \(\hat{\Phi }_x : T_M(x) \rightarrow T_{M'}(\Phi (x))\) defined by \(\hat{\Phi }_x (y) = \Phi (x) + D\Phi _x (y-x)\). Combining Eqs. (1), (2) and (3), we can find a constant \(L_\Phi \) independent of x such that for all \(t < t_0\) and all compact sets \(A \subset M \cap B(x,t)\):
Now, assume that M is nicely triangulable and let us prove that \(M'\) is also nicely triangulable. By definition, we can find \(\rho _0 > 0\) and \(\eta _0>0\) such that, for all \(0 < \rho < \rho _0\), there is a \((\rho ,\eta _0\rho )\)-nice triangulation T of M with respect to some \((h, \mathcal {C})\). Suppose \(\mathcal {C}= \{C_v \mid v \in V\}\) and consider the covering \(\mathcal {C}' = \{ \Phi (C_v) \mid v \in V\}\), the homeomorphism \(h' = \Phi \circ h : |T| \rightarrow M'\), the real numbers \(\rho '=2\kappa _2\rho \) and \(\eta '_0 = \frac{\kappa _1 \eta _0 - 5 L_\Phi \rho }{2\kappa _2}\). Let us prove that by choosing \(\rho \) small enough, T is a \((\rho ',\eta _0'\rho ')\)-nice triangulation of \(M'\) with respect to \((\mathcal {C}',h')\). In other words, we need to check that conditions (ii) and (iii) of Definition 2 are satisfied for \(\mathcal {C}= \mathcal {C}', h=h', \rho =\rho '\) and \(\delta = \eta _0'\rho '\). Take \(v \in V\) and set \(x=h(v), C = C_v, S = h({\text {St}}_{T}(v))\).
(ii) By definition of T, we have \(x \in C \subset {\text {B}}^\circ (x, \rho )\). Taking the image of this relation under \(\Phi \) and choosing \(\rho >0\) small enough, we get that \(\Phi (x) \in \Phi (C) \subset \Phi ({\text {B}}^\circ (x, \rho )) \subset {\text {B}}^\circ (\Phi (x), \rho ')\). The last inclusion is obtained by combining Eqs. (1), (3) and (4).
(iii) Let us choose a positive real number \(\rho < \min { \{ \rho _0, \frac{t_0}{2}\}}\) small enough to ensure that \(\eta _0' > 0\) and let us prove that \(M' \cap [ {\text {Conv}}(C)]^{\oplus \eta _0'\rho '} \subset S\). By choice of T as a \((\rho ,\eta _0\rho )\)-nice triangulation of M with respect to \((h,\mathcal {C})\), we have that \(M \cap {\text {Conv}}(C)^{\oplus \eta _0 \rho } \subset S\). Furthermore, \(C \subset B(x,\rho )\) and \(S \subset B(x,2\rho )\). Thus, by choosing \(\rho < \frac{t_0}{2}\), we have \(S \subset U_x\) and
Taking the image by the homeomorphism \(\pi _x : U_x \rightarrow \pi _x(U_x)\) on both sides and using \(\pi _x(A \cap B) = \pi _x(A) \cap \pi _x(B)\) we get
Let \(B_k(0,r)\) denote the k-dimensional ball of \({\mathcal {T}}_M(x)\) centered at the origin with radius r. Writing \(A\oplus B = \{ a + b \mid a \in A, b \in B\}\) for the Minkowski sum of A and B, it is not too difficult to prove that \(\pi _x(A^{\oplus \delta }) = \pi _x(A) \oplus B_k(0,\delta )\). It follows that
Taking the image under \(\hat{\Phi }_x\) on both sides we get
Let \(B'_k(0,r)\) denote the k-dimensional ball of \({\mathcal {T}}_{M'}(\Phi (x))\) centered at the origin with radius r. Using Eq. (4) we get that \(B_k'(0,\kappa _1\eta _0\rho ) \subset D \Phi _x B_k(0,\eta _0\rho )\). Since \(\hat{\Phi }_x\) and \(\pi _x\) are both affine, so is the composition and therefore \(\hat{\Phi }_x \circ \pi _x({\text {Conv}}(C)) = {\text {Conv}}(\hat{\Phi }_x \circ \pi _x(C))\). It follows that
Recalling that \(C \subset B(x,\rho )\) and \(S \subset B(x,2\rho )\) and combining the above inclusion with Eq. (5) we obtain
Interchanging \({\text {Conv}}\) and \(\pi _x'\), noting that \(\eta '_0\rho ' = \kappa _1 \eta _0 \rho - 5 L_\Phi \rho ^2\) and using \(\pi '_x(A^{\oplus \delta }) = \pi '_x(A) \oplus B_k(0,\delta )\) we get
Since \(\pi '_x : \Phi (U_x) \rightarrow \pi _{\Phi (x)}'(\Phi (U_x))\) is homeomorphic, we thus obtain \(M' \cap {\text {Conv}}(\Phi (C))^{\oplus \eta _0' \rho } \subset \Phi (S)\) as desired. \(\square \)
Rights and permissions
About this article
Cite this article
Attali, D., Lieutier, A. Geometry-driven Collapses for Converting a Čech Complex into a Triangulation of a Nicely Triangulable Shape. Discrete Comput Geom 54, 798–825 (2015). https://doi.org/10.1007/s00454-015-9733-7
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00454-015-9733-7