Scattered data reconstruction by regularization in B-spline and associated wavelet spaces (Q1034074): Difference between revisions

The concern of this paper is the reconstruction of a curve or surface from given scattered data. Scattered data reconstruction (also known as scattered data fitting) problems arise in many fields and applications, such as signal processing, computer graphics and neural networks. The two basic approaches to scattered data reconstruction are interpolation and approximation. A classical approach to scattered data approximation is the solution of the regularized least square problem, where the minimization is taken over all functions belonging to the Beppo-Levi space. The approach of the authors is to solve the minimization problem, not over the Beppo-Levi space, but rather, over the principal shift invariant (PSI) space generated by a single, carefully chosen, compactly supported function. A computational formulation is given in the univariate case when this function is a uniform B-spline and in the bivariate case when this function is the tensor product of uniform B-splines. The PSI space has a simple structure and provides good approximations to smooth functions; it leads to simple and accurate algorithms. The PSI space can be associated to a wavelet system and then one can solve the data fitting problem in the wavelet domain with an efficient algorithm. The proposed method is compared with the classical cubic/thin-plate smoothing spline methods via numerical experiments, where it is seen that the quality of the obtained fitting function is very much equivalent to that of the classical methods, but proposed method offers advantages in terms of numerical efficiency. The numerical experiments demonstrate the computational efficiency of solving data fitting problems in the wavelet domain.