Construction of surfaces by discrete variational splines with parallelism conditions. (Q1426789)

The authors present a new approach for surface fitting in \(\mathbb{R}^2\) which combines properties and techniques of the finite element method (FEM) with those of thin plate splines (TPS). The main theoretical result of the paper consists in obtaining upper estimates of the expected error when the TPSFEM smoother, whose existence is also proved, is applied to uniformly spread data. The TPSFEM procedure is based upon a quasi-uniformity condition of the meshes, and those standard properties which in particular are satisfied by the \(L^2\) projection. The analysis of convergence, which is carried out for a discrete smoothing spline with data perturbed by white noise, is considered by the authors as a novelty. The essential aspects of the numerical method to calculate the TPSFEM smoother can be seen in the article, as well as some experiments to illustrate the power of the method when it is applied to large data sets.