Algorithms for reconstructing a 2-manifold from a point sample in based on Voronoi-filtering like CRUST[1] or CoCone[2] still require --- after identifying a set of candidate triangles --- a so-called manifold extractionstep which identifies a subset of the candidate triangles to form the final reconstruction surface. Non-locality of the latter step is caused by so-called slivers--- configurations of four almost cocircular points having an empty circumsphere with center close to the manifold surface.
We prove that under a certain mild condition --- local uniformity --- which typically holds in practice but can also be enforced theoretically, one can compute a reconstruction using an algorithm whose decisions about the adjacencies of a point only depend on nearby points.
While the theoretical proof requires an extremely high sampling density, our prototype implementation, described in a companion paper [3], preforms well on typical sample sets. Due to its local mode of computation, it might be particularly suited for parallel computing or external memory scenarios.
[1]
Tamal K. Dey,et al.
Manifold reconstruction from point samples
,
2005,
SODA '05.
[2]
Stefan Funke,et al.
How much geometry it takes to reconstruct a 2-manifold in R3
,
2010,
JEAL.
[3]
D. Kirkpatrick,et al.
A Framework for Computational Morphology
,
1985
.
[4]
Mark de Berg,et al.
Computational geometry: algorithms and applications
,
1997
.
[5]
Stefan Funke,et al.
Smooth-surface reconstruction in near-linear time
,
2002,
SODA '02.
[6]
Sunghee Choi,et al.
A Simple Algorithm for Homeomorphic Surface Reconstruction
,
2002,
Int. J. Comput. Geom. Appl..
[7]
Stefan Funke,et al.
Network sketching or: "How Much Geometry Hides in Connectivity?--Part II"
,
2007,
SODA '07.