Regolarizzazione di Tikhonov di problemi mal-posti mediante spline cubiche

In this work Tikhonov regularization with cubic splines for ill-posed problems has been described and applied to the real problem of image restoration. In order to determine the optimum regularization parameter, of this polynomial regularizator, we use a new Monte Carlo Cross-Validation procedure, that permits the use of the Generalized Cross-Validation method with a workload similar to the standard residual method for large least square problems with noisy data. We compute the minimum of the GCV function by the iterative golden section search scheme, that implics the solution of a linear system for each iteration