Integro spline quasi-interpolants and their super convergence (Q2204164)

Both quasi-interpolation and quasi-interpolation with univariate splines are at the centre of approximation theory methods and applications. The splines, formed as piecewise polynomials of a certain order and smoothness, are useful because of their flexibility (compact support or boundedness, convergence rate, recurrence relations for defininig and computing them, etc.) and their accuracy due to local polynomial reproduction. The quasi-interpolants are useful due to their localness, their use of the ``right'' approximation (linear) spaces, their smoothing properties while ``inheriting'' the asymptotic accuracy of the piecewise polynomial spaces, and many other positive properties. This article analyses quasi-interpolation (with splines) which is not in the Schoenberg form (of quasi-Lagrange functions summed against [semi-discretely convolved with] pointswise function evaluation) but with (integral) averages as coefficients to the quasi-Lagrange functions. The averages are computed locally, and the authors provide convergence theorems including super-convergence at the knots, and numerical examples.