Mixed Norm-Based Image Restoration Using Neural Network

In this paper, a novel image restoration model is presented based on the adaptive mixed norm regularization and Hop field neural network. The new error function of this image restoration model combines the L2-norm and L1-norm. To fit the neural network processing, the nonlinear gradient operator of L1-norm is decomposed to the sum of linear operators. Two methods of calculating the adaptive scale control parameter and the modified implementation technique using neural network are presented. Experimental results demonstrate the proposed algorithms are more effective than the traditional and total variation image restoration algorithms.