A shape-preserving approximation by weighted cubic splines (Q442724)

Let \(-\infty < a = x_0 < x_1 < \dots < x_{N+1} = b < \infty\). The shape-preserving interpolation problem consists of constructing a weighted cubic \(C^1\)-spline \(s\) so that \(s(x_j) = f_j\) \((j=0,\dots, N+1)\) and \(s\) is monotonic and convex on the intervals of monotonicity and convexity of the given data \((x_j,f_j)\) \((j=0,\dots, N+1)\). For positive weights \(w_j\) \((j=0,\dots, N+1)\), a weighted cubic \(C^1\)-spline \(s\) is a cubic \(C^1\)-spline which satisfies the conditions \(w_{j-1}\,s''(x_j -0) = w_j\, s''(x_j + 0)\) \((j=1,\dots, N)\). If \(w_j = 1\) for all \(j\), then \(s\) is a cubic \(C^2\)-spline. In this paper, the authors construct weighted \(C^1\)-splines and estimate the interpolation error. The weights \(w_j\) can be chosen depending on the monotonicity and convexity of the data. A weighted cubic B-spline basis is presented and well suited for the generation of freeform curves. Numerical examples for the shape-preserving interpolation of functions and curves are given.