workshop beamer

8/3/2019 Workshop Beamer

1/132

Voronoi Diagrams

Voronoi Diagrams

Swami Sarvottamananda

Ramakrishna Mission Vivekananda University

NIT-IGGA, 2010
http://find/http://goback/


2/132

Voronoi Diagrams

Outline I

1 IntroductionMotivation for Voronoi DiagramHistorical Notes and Web References

2 Voronoi Diagrams ConceptsDefinitionsFormal Definition of Voronoi Diagram

3 Computing Voronoi DiagramsNon-optimal AlgorithmsVoronoi Diagram from 3D Convex Hulls

Fortunes AlgorithmDivide and Conquer

4 Delaunay TriangulationsMotivation

ConceptsMethod of Computation


3/132

Voronoi Diagrams

Outline II

5 GeneralisationsVoronoi Diagrams in d

Voronoi Diagram for Farthest PointsVoronoi Diagram for Subsets

6 ConclusionOpen ProblemsConcluding RemarksFurther Reading


4/132

Voronoi Diagrams

Introduction

Voronoi Diagrams


5/132

Voronoi Diagrams

Introduction

Voronoid Diagram

Example (What does it look like?)

Each cell is a set of points whose nearest site is the same

V i Di


6/132

Voronoi Diagrams

Introduction

Motivation

Applications in Cross-disciplines

This concept has independently emerged, and proven useful, invarious fields of sciences

Medial axis transform in biology and physiology,

Wigner-Seitz zones in chemistry and physics,

Domains of action in crystallography,

and Thiessen polygons in meteorology and geography.

V i Di


7/132

Voronoi Diagrams

Introduction

Motivation

Other Direct ApplicationsI

Example (Jurisdiction of Schools in Boston)

A relevant modern day use.

Voronoi Diagrams


8/132

Voronoi Diagrams

Introduction

Motivation

Other Direct ApplicationsII

Example (1854 Cholera Epidemic in London)

A particularly notable use of a Voronoi diagram was the analysis ofthe 1854 cholera epidemic in London, in which physician JohnSnow determined a strong correlation of deaths with proximity to aparticular (and infected) water pump on Broad Street.

Voronoi Diagrams


9/132

Voronoi Diagrams

Introduction

Motivation

John Snows Voronoi Diagram

Anyone can now guess the infected pump.

Voronoi Diagrams


10/132

Voronoi Diagrams

Introduction

Motivation

Motivation

Problem (Voronoi Diagrams The Post Office Problem)

In a city with several post offices we would like to mark the serviceregion of each post office by proximity. What are these regions?

This problem is mentioned in Chapter 7 of the computationalgeometry book by de Berg et al. [3].

Let us solve the problem for a section of Kolkata.

Voronoi Diagrams


11/132

Voronoi Diagrams

Introduction

Motivation

Post-offices in a section of Kolkata

Courtesy: Google Images

Voronoi Diagrams


12/132

g

Introduction

Motivation

Postoffice Problem and its Simplification

We would like to demarcate the service regions of the post-offices.

To simplify, let us use the Euclidean distance (how the crows fly)and not the actual distance by the roads.Secondly, post offices are considered as points.

Voronoi Diagrams


13/132

g

Introduction

Motivation

Post-offices as Points in Plane

First, we remove the clutter.

Voronoi Diagrams


14/132

Introduction

Motivation

Proximity Regions of Postoffices

Second, we compute the Voronoi Diagram

Voronoi Diagrams


15/132

Introduction

Motivation

Postoffice Services in Kolkata

Last, we superimpose the solution. Gives correct answer for atleast Swami Vivekanandas Birth Place at Simla.

Voronoi Diagrams


16/132

Introduction

History

Voronoi Diagrams in Internet

Where you can play around with Voronoi Diagrams

Java application athttp://www.nirarebakun.com/voro/ehivorocli.html(includeshigher order VDs),

VoroGlide (another Java application)http://wwwpi6.fernuni-hagen.de/GeomLab/VoroGlide/

Anything related to Computational Geometry, start at JeffEricksons web page,

http://compgeom.cs.uiuc.edu/ jeffe/compgeom/compgeom.htmlAnd, much more.

(All the hyperlinks were alive in January 2010, I checkedpersonally)

Voronoi Diagrams


17/132

Introduction

History

Origin of Problem

The problem is centuries old

Earliest record in Principles of Philosophy [4] by R. Descartes,17th century.

Decomposition of space into convex regions of influence ofheavenly bodies.

Voronoi Diagrams


18/132

Introduction

History

Origin of Problem

Example (Descartes Explanation of Gravity)

Of course gravity does not work like this, but Newton and Einstienwere to come later.

Voronoi Diagrams


19/132

Voronoi Diagrams Concepts

Definitions

What are Voronoi Diagrams

What are Voronoi Diagrams?

Voronoi Diagrams

V i Di C


20/132


Definitions

What are Voronoi Diagrams

What are Voronoi Diagrams?

Informally, Voronoi Diagrams are proximity regions to a set ofobjects. We collect all the points in a set for whom, someparticulat proximity query gives the same result.


21/132

Voronoi Diagrams



22/132


Definitions

Voronoid Diagram

Example (Voronoi Diagram in Plane)

Voronoi Diagrams



23/132


Definitions

Proximity Queries

A dataset S of n points called sites in 2.

Let S = {s1, s2, . . . , sn}.

Voronoi Diagrams



24/132


Definitions

Proximity Queries


Let S = {s1, s2, . . . , sn}.(Query 1) For any given query point find the closest site to q.

Voronoi Diagrams



25/132


Definitions

Proximity Queries



(Query 2) Find the proximity region of a site.

Voronoi Diagrams



26/132

g p

Definitions

Proximity Queries



(Query 2) Find the proximity region of a site.

(Query 3) Find the proximity regions of all sites.

Voronoi Diagrams



27/132

g p

Definitions

Proximity Query Illustration

Example (Sites and Query Point)

What is nearest point to q?

Voronoi Diagrams



28/132

Definitions

Proximity Query Illustration

Example (Nearest Site to Query Point)

Also, circle of shortest radius is empty of other sites.

Voronoi Diagrams



29/132

Definitions

Proximity Query Easy Solution

Naturally the solution can be found in linear time.Just calculate the distances from all sites and choose the

closest site.


30/132


31/132


32/132

Voronoi DiagramsVoronoi Diagrams Concepts


33/132

Formal Definition

Computing Voronoi Diagram

Example (What is Voronoi Diagram of 2 points)

s1

s2

Closest to s2!!Closest to s1!!

V(s2)?

V(s1)?

Partition the plane in two sets. Points nearer to s1 and nearer to s2


F l D fi i i


34/132

Formal Definition

Computing Voronoi Diagram

Example (Voronoi Diagram of 2 points)

s1

s2

Closest to s2Closest to s1

V(s2)

V(s1)

Voronoi edge

Its perpendicular bisector


F l D fi iti


35/132

Formal Definition

Computing Voronoi Diagram of 3 points

Example (What is Voronoi Diagram of 3 points)

s1

s2

s3

V(s3)?

V(s1)?

V(s2)?

Voronoi vertex

Perpendicular Bisector(s1,s2)

Partition the plane in three sets. Points nearer to s1, s2 and s3


36/132


37/132


Formal Definition


38/132

Formal Definition

Properties

Voronoi edges are in part perpendicular bisectors.

Voronoi vertices are equidistant to three points(circumcentre).


Formal Definition


39/132

Formal Definition

Properties

Voronoi edges are in part perpendicular bisectors.

Voronoi vertices are equidistant to three points(circumcentre).

Not immediately obvious, the circumcircle must be empty(Let us see this again).


Formal Definition


40/132

Properties of Voronoi Diagram

Example (There can not be a point nearer than the three points)

s1

s2

s3

V(s3)

V(s1)

V(s2)

Voronoi vertex

Perpendicular Bisector

Circumcircle is empty and voronoi vertex in centre of circumcircle.This motivates an algorithm.

Voronoi DiagramsComputing Voronoi Diagrams

Non-optimal Algorithms


41/132

p g

Straight-forward Algorithm

Frontal Attack to the Problem

Computen3

circumcentres

Remove non-empty circumcircles

Intelligently join circumcentres by bisectors (whenever twopoints are common)

Voila, you get the Voronoi Diagram

Bad algorithm, O(n4) time complexity




42/132

Space Complexity of Voronoi Diagrams

Before giving good algorithms, we ask what is the size of output.Why?Because if output is big we do not hope to improve.


43/132


44/132


45/132




46/132

Proof of Space Complexity

Eulers relation for planar graph (Delaunays) is v + f = e+ 2


47/132




48/132



e 3v 6, again from Euler,and therefore, f 2v 4




49/132



e 3v 6, again from Euler,and therefore, f 2v 4

Since there is an implicit face at , which corresponds to novoronoi vertex, number of voronoi vertices f 2v 5 .




50/132

Another Algorithm

Each Voronoi cell is intersection of n 1 half planes

We need to compute intersections efficiently

Voronoi Diagrams

Computing Voronoi Diagrams



51/132

Another AlgorithmAnalysis

Each V(si) is intersection of n 1 half planes

Voronoi Diagrams




52/132



Intersection of n halfplanes can be computed in O(n log n)time


53/132

Voronoi Diagrams




54/132



Intersection of n halfplanes can be computed in O(n log n)time

Voronoi diagram of S can be computed in O(n2 log n) time bycomputing the Voronoi cells one by one.

A large improvement over previous method (from n4)

Voronoi Diagrams




55/132

Yet Another Algorithm (Incremental)

Each point is inserted in Voronoi Diagram one by one

A

B C

D

N


56/132

Voronoi Diagrams




57/132

Yet Another Algorithm (Analysis)

Site si+1 is located in Voronoi Diagram of i sites (by rayshooting?)

Its boundary is calculated

Voronoi Diagrams




58/132




Voronoi diagram of S can be computed in O(n2) as O(i)edges might be inserted or deleted in each step

Voronoi Diagrams



( )


59/132




Voronoi diagram of S can be computed in O(n2) as O(i)edges might be inserted or deleted in each step

Further improvement over previous method

Voronoi Diagrams



O


60/132

One more Incremental Algorithm-I

Each Voronoi vertex is inserted in Voronoi Diagram one by one

We start with any hull edge and compute the empty circle

Voronoi Diagrams



O I l Al i h II


61/132

One more Incremental Algorithm-II

Compute adjacent empty circles

Voronoi Diagrams



O I l Al i h III


62/132

One more Incremental Algorithm-III

Final Voronoi Diagram


63/132


64/132

Voronoi Diagrams



Th ee Po la Algo ith s


65/132

Three Popular Algorithms

We know of three good algorithms.

Fortunes Sweep Line (Beach Line) MethodDivide and Conquer

Reduction to Convex Hull in 3

Voronoi Diagrams


Voronoi Diagram from 3D Convex Hulls



66/132


Transform points in plane to space

(x, y) (x, y, x2

+ y2

)Basically we project the plane to the surface of a paraboloidThe lower convex hull is Delaunay Triangulation, the dual ofVoronoi Diagram

Voronoi Diagrams


Voronoi Diagram from 3D Convex Hulls

Projection to Paraboloid


67/132

Projection to Paraboloid

Reduction to Convex Hull

Lower Convex Hull is Delaunay Triangulation.

Voronoi Diagrams


Fortunes Algorithm


68/132

Fortunes Algorithm

Voronoi Diagrams


Fortunes Algorithm

Assumptions


69/132

Assumptions

General position assumption: No four sites are co-circular.

Figures sources:http://www.ams.org/featurecolumn/archive/voronoi.html

Voronoi Diagrams


Fortunes Algorithm

Why Usual Sweep Line Methods Fail?


70/132

Why Usual Sweep Line Methods Fail?

The problem with usual sweeping methods is that the cells maystart even before the site is encountered.

Voronoi Diagrams


Fortunes Algorithm

Tentative Idea


71/132

Tentative Idea

Fortunes algorithm is sweep line method [6]

s2

s1

s3

beach line

sweep line

Known

Unknown

Region inside beach line does not depend on anything below thesweep line. What is it?

Voronoi Diagrams


Fortunes Algorithm

Tentative Idea


72/132

Tentative Idea

Fortunes algorithm is sweep line method [6]

s2

s1

s3

beach line

sweep line

Known

Unknown

Beach line consists of sequence of parabolas. Why?


73/132

Voronoi Diagrams


Fortunes Algorithm

What is Beach line-I


74/132

What is Beach line I

It is parabola.We know equidistant points from a site and a line is a parabola.Anything in the parabola is nearer to the point always.

Voronoi Diagrams


Fortunes Algorithm

What is Beach line-II


75/132

What if we have more points?

Suppose we have got six sites

Voronoi Diagrams


Fortunes Algorithm

What is Beach line-III


76/132

Three sites are already seen by sweep line. Another three are yet tobe seen.

Voronoi Diagrams


Fortunes Algorithm

What is Beach line-IV


77/132

Beach line will be the envelope of the three parabolae

Voronoi Diagrams


Fortunes Algorithm

Important Property of Beach Line


78/132

p p y

We can order the parabolic segments from left to right (or right to

left)

Beach line intersects any vertical line only once

Voronoi Diagrams


Fortunes Algorithm

Two types of Events


79/132

Sweep line can have two types of events

Site Event (Voronoi Edges)

Circle Event (Voronoi Vertices)

The algorithm basically keeps track of the beach line, whilecomputing Voronoi Diagram inside it.

Voronoi Diagrams


Fortunes Algorithm

Initial Situation


80/132

Voronoi Diagrams


Fortunes Algorithm

A site is encountered


81/132

Voronoi Diagrams


Fortunes Algorithm

Voronoi Edge Enters


82/132

That means we have a Voronoi edge whenever two beach segmentsintersect

Voronoi Diagrams


Fortunes Algorithm

Voronoi Vertex Enters


83/132

There is a Voronoi vertex whenever two Voronoi edges intersectWe need to add an event (circle event) where three neighbouringbeach segments intersect

Voronoi Diagrams


Fortunes Algorithm

Circle Event-I


84/132

Before circle event

Voronoi Diagrams


Fortunes Algorithm

Circle Event-II


85/132

At the circle event

Voronoi Diagrams


Fortunes Algorithm

Circle Event-III


86/132

After circle event

Voronoi Diagrams


Fortunes Algorithm

Execution of the Fortunes Algorithm


87/132

Let us see how the Fortunes algorithm calculates the VoronoiDiagram step by step

Voronoi Diagrams


Fortunes Algorithm

AlgorithmStep 1


88/132

First site and a strange beach line

Voronoi Diagrams


Fortunes Algorithm

AlgorithmStep 2


89/132

Second site


90/132

Voronoi Diagrams


Fortunes Algorithm

AlgorithmStep 4


91/132

Fourth site (and an effective circle event)

Voronoi Diagrams


Fortunes Algorithm

AlgorithmStep 5


92/132

Circle event

Voronoi Diagrams

Computing Voronoi DiagramsFortunes Algorithm

AlgorithmStep 6


93/132

Fifth site and a circle event

Voronoi Diagrams


AlgorithmStep 7


94/132

A circle event and another circle event


95/132

Voronoi Diagrams


AlgorithmStep 9


96/132

Last site and two circle event

Voronoi Diagrams


AlgorithmStep 10


97/132

Last but one circle event

Voronoi Diagrams


AlgorithmStep 11


98/132

Last circle event

Voronoi Diagrams


AlgorithmStep 12


99/132

Final Output

Voronoi Diagrams


Analysis of the Fortunes Algorithm


100/132

Fortunes algorithm being an example of typical sweep linetechnique is O(n log n)

Optimal because sorting problem can be reduced to construction ofVoronoi Diagrams

Voronoi Diagrams


Reduction of Sorting Problem to Voronoi Diagram


101/132

To sort x1, x2, . . . , xnWe find Voronoi Diagram of points(x1, 0), (x2, 0), . . . (xn, 0) and (0,) = sEither neighbourhood relation or the cell V(s) gives the requiredsorted order

Voronoi Diagrams


Sorting using Voronoi Diagrams


102/132

Voronoi Diagrams

Computing Voronoi DiagramsDivide and Conquer

Divide and Conquer


103/132

Another algorithm that gives Voronoi Diagram in optimal time andspace

Voronoi Diagrams


Divide


104/132

Partition points equally. Solve two subproblems.

Voronoi Diagrams


Conquer


105/132

Sew the two jumbled pieces

Voronoi Diagrams


Solution


106/132

Merging of Voronoi Diagram complete!!

Voronoi Diagrams

Delaunay TriangulationsMotivation


107/132

Delaunay Triangulations

Voronoi Diagrams


What are Delaunay Triangulation?


108/132

Given a set of points sometimes it is needed to join them intriangles

Voronoi Diagrams


Bad triangulation


109/132

Example (Skinny Triangles)

Voronoi Diagrams


Good triangulation


110/132

Example (Delaunay Triangulation)

Voronoi Diagrams


Concepts

Delaunay triangulation as Dual of Voronoi Diagrams


111/132

Example (Delaunay Triangles are Voronoi Vertices)

Voronoi Diagrams


Concepts

Another Example to Convince


112/132

Example (Delaunay Triangulation of 50 points)

Voronoi Diagrams


Concepts

Yet Another Example to Convince


113/132

Example (Delaunay Triangulation of 100 points)

Voronoi Diagrams


Concepts

What are we looking for?


114/132

Triangles should not be very stretched (skinny triangles)

Maximize the minimum angle of all the angles of the trianglesin the triangulation

Circumcircles of all the triangles in the net are empty

Voronoi Diagrams


Concepts

Delaunay Triangulation Properties-I


115/132

Example (Delaunay Triangles have Empty Circumcircles)

Voronoi Diagrams


Concepts

Delaunay Triangulation Properties-II


116/132

Example (Empty Circumcircles centred at Voronoi Vertices)

Voronoi Diagrams


Method of Computation

Delaunay Triangulation Properties-II


117/132

Example (Empty CircumcirclesDual of Voronoi Diagrams)

Voronoi Diagrams

Generalisations


118/132

Generalisations

Voronoi Diagrams

Generalisations

Voronoi Diagrams in d

Higher Dimension Voronoi Diagram

Example (Voronoi Region 3D point set)


119/132

Voronoi Diagram in space

Voronoi Diagrams

Generalisations

Voronoi Diagram for Farthest Points

Farthest Point Voronoi Diagram


120/132

Example (Voronoi regions for farthest point queries)

Voronoi Diagrams

Generalisations


Computing Farthest Point Voronoi Diagram


121/132

Farthest Point Voronoi diagrams only depends on points onConvex Hull .

This makes the task of computing the Farthest Point VoronoiDiagram easier and gives rise to O(n log n) algorithm.

Voronoi Diagrams

Generalisations


Applications of Farthest Point Voronoi Diagram


122/132

Used for Min-Max optimisation problems

Gives you minimum enclosing circle directly. Though linear timealgorithms that do not use farthest point Voronoi Diagram areknown.

Voronoi Diagrams

Generalisations

Voronoi Diagram for Subsets

Higher Order Voronoi Diagram

Example (Voronoi Region for k-nearest point set)


123/132

Second order Voronoi Diagram

Voronoi Diagrams

Generalisations

Voronoi Diagram for Subsets

Computing Higher Order Voronoi Diagram


124/132

Higher-order Voronoi diagrams can be generated recursively.

To generate the nth-order Voronoi diagram from set S, start with

the (n 1)th

-order diagram and replace each cell generated byX = {x1, x2, ..., xn1} with a Voronoi diagram generated on the setS X.

Collect all the regions of X in different cells.

Voronoi Diagrams

Conclusion


125/132

Conclusion

Voronoi Diagrams

Conclusion

Open Problems

Open Problems to Ponder


126/132

Compute Voronoi diagram of a set of lines (or line segments)in three dimensions, conjectured to be near quadratic. (Samefor Polyhedra is hard)

What is the maximum number of combinatorial changespossible in a Euclidean Voronoi diagram of a set of n pointseach moving along a line at unit speed in two dimensions?

Where to place a new site to maximize its Voronoi cell?

Voronoi Diagrams

Conclusion

Concluding Remarks

Summary


127/132

We studied the concept of Voronoi Diagrams

Next we saw how they can be computed

We looked into their dualsDelaunay Triangulations

Finally we also saw some of the generalisations

Voronoi Diagrams

Conclusion

Further Reading

You may read


128/132

Book by Okabe et al. [7], who lists more than 600 papers, and

the surveys by Aurenhammer [1],

Bernal [2], andFortune [6] for a complete overview,

Also, Chapters 5 and 6 of Preparata and Shamos [8], and

Chapter 13 of Edelsbrunner [5]

Voronoi Diagrams

Conclusion

Further Reading

References I

Franz Aurenhammer.

Voronoi diagrams a s r e of a f ndamental geometric data


129/132

Voronoi diagrams - a survey of a fundamental geometric datastructure.ACM Comput. Surv., 23(3):345405, 1991.

Javier Bernal.

Bibliographic notes on voronoi diagrams.Technical report, Natl. Inst. of Standards and Tech., April1993.NIST Internal Report 5164.

Mark de Berg, Marc van Kreveld, Mark Overmars, and OtfriedSchwarzkopf.Computational Geometry: Algorithms and Applications.Springer-Verlag, second edition, 2000.

Voronoi Diagrams

Conclusion

Further Reading

References II


130/132

Rene Descartes.Principia Philosophiae.Ludovicus Elzevirius, Amsterdam, 1644.

Herbert Edelsbrunner.Algorithms in Combinatorial Geometry.Springer-Verlag, New York, 1987.

Steven Fortune.Voronoi diagrams and Delaunay triangulations.

CRC Press, Inc., Boca Raton, FL, USA, 1997.

Voronoi Diagrams

Conclusion

Further Reading

References III


131/132

Atsuyuki Okabe, Barry Boots, Kokichi Sugihara, andSung Nok Chiu.Spatial tessellations: Concepts and applications of Voronoidiagrams.

Probability and Statistics. Wiley, NYC, 2nd edition, 2000.671 pages.

F. P. Preparata and M. I. Shamos.Computational Geometry: An Introduction.

Springer-Verlag, 1985.

Voronoi Diagrams

Conclusion

Further Reading

At Last . . .


132/132

Thank You

[email protected]

[email protected]

workshop beamer

Documents