A non-uniform corner-cutting subdivision scheme with an improved accuracy (Q2656098)

Corner cutting and subdivision algorithms are methods for computing (spline- or otherwise) approximations of surfaces in efficient and stable recursive schemes. The coefficients of these recursive algorithms are called masks. They can result in piecewise polynomial surfaces or for instance piecewise exponentials. The schemes can be formed with equally spaced or non-meshed data points (knots). In this article the non-stationary case for piecewise exponentials is considered. The particular focus is not only on the efficiency of the algorithm but on its asymptotic accurracy, achieved by reproducing exponential functions, and especially the improvement of the convergence order from the results of the classical schemes. The essential parts of the work contain (i) a suitable choice of the so-called subdivision mask, i.e., the coefficients, (ii) using the correct parameters for the exponentials (i.e., their coefficients in the exponent), and (iii) deriving convergence results from this. For (ii), the reproduction of the exponentials that can later-on be used in the asymptotic convergece estimates is particularly important, but the authors noted that only local reproduction of exponentials is needed because the convergence results are local too. For (iii), Theorem~5.6 is the most relevant result. The authors give several numerical examples at the end of the article as well.