Adaptive image restoration using a multiresolution Hopfield neural network

The paper describes an approach to the restoration problem which combines the energy minimization of Hopfield networks with a multiresolution image representation, which allows coarse-fine smoothing, in a way similar to that pioneered by Clippingdale (1988). After a brief presentation of the theoretical background, results are presented for an adaptive restoration algorithm and compared with those obtained using a non-adaptive technique, which show that the adaptive method gives significantly better results, in terms of their subjective quality