Surface fitting and compression of two-dimensional scattered data

The problem of efficient smoothing and compression of large sets of two-dimensional data is addressed. Global surface fitting methods using radial basis functions are shown to have inherent characteristics which enable a simplification of the general nonlinear problem. An efficient placement of basis functions reduces their number considerably. The separate calculation of basis function locations leaves only the problem of evaluating their linear coefficients. A least-square method can be used, but for large sets of data, a further relaxation using information gained during the placement process, enables the application of linear programming techniques. A few examples of fitting topographical data are given