i
i
i
i
i
i
i
i
364 15. Curves
12
3
4
5
67
Figure 15.8. Cardinal splines through seven control points with varying values of tension
parameter
t
.
Cardinal splines are useful, because they provide an easy way to interpolate
a set of points with C
1
continuity and local control. They are only C
1
,sothey
sometimes get “kinks” in them. The tension parameter gives some control over
what happens between the interpolated points, as shown in Figure 15.8, where a
set of cardinal splines through a set of points is shown. The curves use the same
control points, but they use different values for the tension parameters. Note that
the first and last control points are not interpolated.
Given a set of n points to interpolate, you might wonder why we might prefer
to use a cardinal cubic spline (that is a set of n −2 cubic pieces) rather than a sin-
gle, order n polynomial as described in Section 15.3.6. Some of the disadvantages
of the interpolating polynomial are:
• The interpolating polynomial tends to overshoot the points, as seen in Fig-
ure 15.9. This overshooting gets worse as the number of points grows
larger. The cardinal splines tend to be well behaved in between the points.
• Control of the interpolating polynomial is not local. Changing a point at the
beginning of the spline affects the entire spline. Cardinal splines are local:
any place on the spline is affected by its four neighboring points at most.
• Evaluation of the interpolating polynomial is not local. Evaluating a point
on the polynomial requires access to all of its points. Evaluating a point
on the piecewise cubic requires a fixed small number of computations, no
matter how large the total number of points is.
There are a variety of other numerical and technical issues in using interpolating
splines as the number of points grows larger. See (De Boor, 2001) for more
information.
A cardinal spline has the disadvantage that it does not interpolate the first or
last point, which can be easily fixed by adding an extra point at either end of