Scribd
Upload
37 views

A Comparison of Three Curve Intersection Algorithms

This document presents an empirical comparison of three algorithms for finding intersection points of planar Bezier curves: a well-known Bezier subdivision algorithm, an interval analysis-based algorithm, and a polynomial root-finding algorithm. The results indicate that the polynomial root-finding algorithm is significantly faster, especially for multiple intersections, while the performance of the other algorithms varies with the degree of the curves. The study also highlights the importance of implementation choices and termination criteria in affecting execution speed.

Uploaded by

bim.frankliang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
37 views

A Comparison of Three Curve Intersection Algorithms

This document presents an empirical comparison of three algorithms for finding intersection points of planar Bezier curves: a well-known Bezier subdivision algorithm, an interval analysis-based algorithm, and a polynomial root-finding algorithm. The results indicate that the polynomial root-finding algorithm is significantly faster, especially for multiple intersections, while the performance of the other algorithms varies with the degree of the curves. The study also highlights the importance of implementation choices and termination criteria in affecting execution speed.

Uploaded by

bim.frankliang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

A Comparison of

Three Curve Intersection Algorithms

Thomas W. Sederberg
Scott R. Parry
Brigham Young University
Provo, Utah 84602

EXTENDED ABSTRACT
An empirical comparison is made between three algorithms for computing
the points of intersection of two planar Bezier curves. The algorithms compared
are:
Al. The well known Bezier subdivision algorithm, which is discussed in [Lane180]
A2. A subdivision algorithm based on interval analysis due to Koparkar and
Mudur [Koparkar183]
A3. An algorithm due to Sederberg, Anderson and Goldman which reduces the
problem to one of finding the roots of a univariate polynomial [Seder-
berg'84).
The details of these three algorithms are presented in their respective references.
This abstract mentions major implementation choices, and assumes that the
reader is familiar with the three algorithms.
The execution times of A1 and A2 are highly sensitive to implementation.
We observed that the decision on when to terminate the subdivision can effect
execution speed by an order of magnitude. We based our termination criterion
on a suggestion from [Wang'84] that a curve be subdivided until its deviation
from a straight line segment is no larger than a value E . If (xi,yi) i = O , l , ...,n are
the control points of a degree n Bezier curve and

then

where ro is the number of times the curve must be subdivided in order for a
curve segment to approximate a line segment to within a tolerance c.
For A1 and A2, the convex hull was taken to be a bounding rectangle. For
A l , subdivision was performed using the geometric construction algorithm and
the bounding rectangle was determined by the new Bezier control points. For
A2, the polynomial equations were converted to standard power basis, and subdi-
vision was performed by simply evaluating the curve segment at its parametric
midpoint, using Horner's algorithm.
For A3, the implicitization procedure was performed using Bernstein polyno-
mials. It was noted that standard polynomials introduced excessive roundoff
error for curves of degree larger than three. A critical part of the implementation
is a polynomial root finder which finds only real roots. We developed our own
root finder for this purpose, which usually runs significantly faster than, for
example, the standard IMSL root finder which finds all roots (real and complex)
of a polynomial.
We tested these three algorithms on pairs of Bezier curves of degree three,
four, and five, with several different numbers of intersection points. Our goal was
to compute the intersection points to five decimal places of accuracy, and we
experimented to find the optimal value of ro for this purpose.
Results: The tests were run on a VAX11/750, with floating point accelera-
tor, under UNlX 4.2 BSD. Our test results indicate that for two cubic Bezier
curves which intersect once, A3 is at least twice as fast as the other two algo-
rithms. For more than one intersection, A3 does even better -- if the two cubic
curves intersect nine times, A3 is typically five or six times faster than the others.
For curves of degree four, A2 and A3 are comparable, and A1 is the slowest.
For curves of degree five, A2 is typically twice as fast as the other two algo-
rithms.
Multiple intersections were not thoroughly tested. Our implementation of
A1 and A2 did not properly detect the presence of tangency. A3 handled
tangency with no problem by reporting a multiple polynomial root.

References
Koparkar, P.A. and Mudur, S.P.(1983), "A new class of algorithms for process-
ing -parametric curves", Computer-Aided Design, Vol. 15, No. 1, 41-
45.
Lane, J.M., and Riesenfeld, R.F.(1980), "A theoretical Development for the
Computer Generation and Display of Piecewise Polynomial Sur-
faces", IEEE Transactions on Pattern Analysis and Machine Intelli-
gence, Vol. PAMI, No. 1,35-46.
Sederberg, T.W., Anderson, D.C. and Goldman, R.N.(1984), "Implicit
representation of parametric curves and surfaces", Conaputer
Vision, Graphics and Image Processing, Vol. 28, 72-84.
Wang, G. (1984), "The Subdivision Method for Finding the Intersection
between Two Bezier Curves or Surfaces", Zhejiang University Jour-
nal, Special Issue on Computational Geometry.

You might also like