Bivariate interpolation based on univariate subdivision schemes (Q420770): Difference between revisions

Let \(N\in\mathbb{N}\) and \(\triangle:a=x_{0}<x_{1}<\dots <x_{N}=b\) be a division of the interval \([a,b].\) Consider the rectangular domain \(\Omega =[a,b]\times[-\pi,\pi]\subset\mathbb{R}^{2}\) and the data functions \(f_{i}(y)\) given along the partition line \(x=x_{i},\) \(i=0,1,2,\dots,N\). The authors solve the following problem: Find a smooth bi-variate interpolant \(u:\Omega\rightarrow\mathbb{R}\) satisfying \(u(x_{i},y)=f_{i}(y),\) \(y\in [-\pi,\pi],\) \(i=0,1,2,\dots,N\). This problem is treated for an equally-spaced partition \(\triangle\) and \(f_{i}(y)\) being a \(2\pi\)-periodic function, for \(i=0,1,2,\dots,N\). One obtains an interpolant of the form \(u(x,y)=\sum_{m=0}^{\infty}(c_{m}(x)\cos(my)+s_{m}(x)\sin(my)),\) where the coefficients \(c_{m}(x)\) and \(s_{m}(x)\) are limits of non-stationary 4-point subdivision schemes \(S_{m}\), see \textit{N. Dyn, D. Levin} and \textit{J. Gregory} [Comput. Aided Geom. Des. 4, 256--268 (1987; Zbl 0638.65009)], obtained from the initial data \(c_{m}^{0}=\{c_{i,m}\}_{i=0}^{N},\) \(s_{m}^{0}=\{s_{i,m}\}_{i=0}^{N},\) \(m\in\mathbb{Z}_{+}.\) Some properties of \(u(x,y)\) are derived, such as continuity, smoothness and approximation order.