Image restoration by Hopfield networks considering the line process

A technique for the restoration of an image degraded by a shift invariant point spread function (SIPSF), and corrupted with white Gaussian noise is developed. For taking account of the effect of edges in an image, two kinds of Hopfield networks are used alternately in the proposed technique. One is for the edge detection in images and another is for the restoration and smoothing of blurred images. For exact edge detection, a Hopfield network considering the line process is proposed. To achieve the fast convergence of optimizing process of restoration and smoothing network, the regularization parameter is chosen to be a sigmoid function in which its iteration number is considered as a variable. An example is shown to illustrate the utility of the proposed technique