On the Regularization of Ill-Posed Problems