Post on 30-Dec-2015
PRE-TRIANGULATIONS Generalized Delaunay Triangulations and Flips
Franz Aurenhammer
Institute for Theoretical Computer Science
Graz University of Technology, Austria
Why do we like Voronoi diagrams?
What do we do when a (nice) structure does not exactly fit our purposes?
Generalize
- Shape of sites- Distance function- Underlying space
Hand in hand with Voronoi diagrams goes the Delaunay triangulation
Surprisingly, duals of generalized Voronoi diagrams play a minor role
Why are Delaunay triangulations harder to generalize than Voronoi diagrams?
Voronoi diagram: Fix properties (mainly distance function), study the shape of regions
Delaunay triangulation: Fix the shape of regions (triangles), study resulting (combinatorial) properties.
Generalize the Delaunay triangulation independently!
What are Delaunay triangulations special for?
- Unique structure
- Local ‘Delaunayhood‘
- Flippability of edges
- Liftability to a surface in 3D
When generalizing the Delaunay triangulation, we want to keep these properties.
How to generalize a triangulation, anyway?
Triangle: Exactly 3 vertices without reflex angle
Pseudo-triangle, Pre-triangle
Pseudo-triangulations
Pre-triangulations
Data Structure: Visibility, collision detection
Graph: Rigidity properties
Fairly new concept
Robust liftability of polygonal partitionsis an exclusive privilege of pre-triangulations
How to get ‘Delaunay‘ in …
View the Delaunay triangulationas follows:
S underlying set of points
f* maximal locally convex function on conv(S) such that f*(p) =< p,p> for all p in S
Here: f* is just the lower convex hull
Delaunay Minimum Complex
Restrict values of f* onlyat the corners of thedomain (no reflex angle)
Pseudo-triangulation Unique, liftable, and locally Delaunay (convex)
….not to be confused with the constrained Delaunay triangulation
Delaunay Minimum Complex
Pre-triangulation
Complex of smallest combinatorial size with the desired Delaunay properties!
… and Flippability?
- We should be able to flip any given pre-triangulation into the Delaunay minimum complex
- And flips should be consistent with existing flips for triangulations and pseudo-triangulations
A General Flipping Scheme
FLIP(edge)Choose domain
Give heightsReplace by f*
Flipping Domain
ok no pre-triangulation!
Implications
- Canonical Delaunay pre-triangulation (or pseudo-triangulation) for polygonal regions exists
- Can be reached by improving flips (convexifying flips) from every pre-triangulation
- Extends the well-known properties of Delaunay triangulations
Can we obtain similar results for 3-space?
‘Delaunay‘ for a Nonconvex Polytope
Pseudo-tetrahedra (4 corners)
Bistellar Flip for Tetrahedra
Generalizes for pseudo-tetrahedra!
Exhibition of
(Pseudo)-Delaunay
Art
-simple removing flip-
-simple exchanging flip-
-Splitting off a secondary cell-
-Inserting flip-
-large exchanging flip-
-tunnel flip-
-Thank you-