Adaptive Sub-Optimal Hopfield Neural Network image restoration base on edge detection

In this paper, an adaptive Sub-Optimization Hopfield Neural Network for regularized image restoration based on edge detection is presented. The conventional Hopfield Neural Network restoration is an optimal scheme based on the whole image by minimizing the energy function, which costs too much memory resource since the big `blur matrix'. A large image would cause the computer to run out of memory due to the big `blur matrix'; that means, this optimal scheme cannot restore large size images. To avoid this problem, a Sub-Optimal scheme is utilized that restores the pixels of the distorted image one by one, based on the information of its neighborhood. Adaptive regularization is implemented by using the edge information detected by the Sobel Detector, which can preserve the details and improve the performance of the algorithm. Simulations show the efficiency of the proposed algorithm.