Cubic spline interpolation with quasiminimal B-spline coefficients (Q812822)

There are various concepts of minimization in connection with univariate B-splines and in this paper a very specific one is introduced. Instead of minimizing, for instance, square integrals of certain derivatives of the spline, the sum of squares of coefficients of the B-spline are minimized. This is useful especially for the numerical stability when spline approximations are computed. The degrees of freedom which are needed to this purpose are taken from the end conditions of the spline. In addition to the mentioned minimization, the spline (here a piecewise cubic) is forced to be such that all cubic polynomials are reproduced by the approximation. A few numerical experiments are provided to substantiate the theory.