EVALUATION OF A DELAUNAY-BASED METHOD FOR SURFACE APPROXIMATION

The problem of constructing a bivariate function that approximates a set of data points occurs in a number of important applications. A method (the point-selection method) is presented that constructs a function that interpolates a subset of the data points, and for which the deviation from the remaining points satisfies a given error tolerance. The construction is based on the Delaunay triangulation. It is desirable that the subset interpolated by the constructed function should be as small as possible, that is, that the function should provide an efficient representation of the data. An evaluation of the point-selection method with respect to the efficiency of representation is presented.