An algorithm for computing shape-preserving interpolating splines of arbitrary degree (Q1104031)

A large number of papers dealing with the construction of shape preserving (monotone and/or convex) smooth piecewise polynomial functions (splines) which interpolate data at the knots have appeared in recent years [see, for example, \textit{C. de Boor}, A practical guide to splines (1978; Zbl 0406.41003), \textit{U. Hornung}, J. Approx. Theory 28, 227-237 (1980; Zbl 0432.41001); \textit{D. F. McAllister} and \textit{J. A. Roulier}, Math. Comput. 32, 1154-1162 (1978; Zbl 0398.41004); \textit{D. F. McAllister}, \textit{E. Passow} and \textit{J. A. Roulier}, Math. Comput. 31, 717-725 (1977; Zbl 0371.65001); \textit{H. Mettke}, J. Comput. Appl. Math. 9, 205-211 (1983; Zbl 0523.65006) and by \textit{E. Passow} and \textit{J. A. Roulier}, SIAM J. Numer. Anal. 14, 904-909 (1977; Zbl 0378.41002)]. Quadratic \(C^ 1\) comonotone (CM) and co-convex (CC) interpolating or osculatory splines were constructed by \textit{D. F. McAllister} and \textit{J. A. Roulier} [ACM Trans. Math. Software 7, 331-347 (1981; Zbl 0464.65003)] and by \textit{L. L. Schumaker} [SIAM J. Numer. Anal. 20, 854-864 (1983; Zbl 0521.65009)] by introducing auxiliary knots. A method for monotone and convex splines of degree m was proposed by McAllister, Passow and Roulier (loc. cit.). The author considers (CM) or (CC) or (CM) and (CC) \(C^ k\) splines of degree m where \(0<k<m-k\) and using his results [Math. Comput. 46, 203-214 (1986; Zbl 0617.41015)] an algorithm for their computation [cf. the author and \textit{R. Morandi}, Calcolo 21, 281-294 (1984; Zbl 0565.41006)] (for \(m=3\), \(k=1\) and for monotone and convex interpolation)] is described. The splines considered here are obtained using Bernstein polynomials of suitable continuous piecewise linear functions. In contrast to McAllister, Passow and Roulier (loc. cit.) the algorithm presented here is reasonably simple and fast and the degree m can be chosen a priori or can be automatically computed, according to the sharpness of data. Graphical examples and a listing of the FORTRAN code SPISP1 (Shape-Preserving Interpolating Spline) are also given.