Minimal roughness property of the Delaunay triangulation

A set of scattered data in the plane consists of function values measured on a set of data points in R2. A surface model of this set may be obtained by triangulating the set of data points and constructing the Piecewise Linear Interpolating Surface (PLIS) to the given function values. The PLIS is combined of planar triangular facets with vertices at the data points. The roughness measure of a PLIS is the L2 norm squared of the gradient of the piecewise linear surface, integrated over the triangulated region and obviously depends on the specific triangulation. In this paper we prove that the Delaunay triangulation of the data points minimizes the roughness measure of a PLIS, for any fixed set of function values. This Theorem connects for the first time, as far as we know, the geometry of the Delaunay triangulation with the properties of the PLIS defined over it.