next up previous
Next: About this document Up: Interpolation Previous: Trig interpolation

Splines

Consider now a set of points (knots) tex2html_wrap_inline165 and polynomials defined on each subinterval, tex2html_wrap_inline167 . We think of a function tex2html_wrap_inline169 on the appropriate subintervals. A spline is a piecewise kth degree polynomial that is smoothly connected at each knot.

We shall examine cubic splines, the third order interplation allowing a change in curvature in a given subinterval. We demand

displaymath171

There are n polynomials, so there are 4n coefficients that need to be determined. These constraints give n+1 + 3(n-1) = 4n-2 conditions. The extra 2 conditions are usually specified as derivative information at the ends. Free splines specify zero second derivatives at btoh ends; clamped splines specify the first derivatives at the ends.

On tex2html_wrap_inline173 , consider the spline S. Because it is cubic, its second derivative is linear, tex2html_wrap_inline175 , and we have

eqnarray71

By evaluating at the two knots, we find

displaymath177

Now impose the continuity of deivatives from left and right. After considerable algebra, we find

displaymath179

This is n-1 equations for the n+1 unknowns tex2html_wrap_inline181 . A free spline implies tex2html_wrap_inline183 , while a clamped spline gives derivative conditions for S'. We are left with a tridiagonal system to solve.



E. Bruce Pitman
Fri Nov 27 18:33:05 EST 1998