A Neural Network Based Solution to Color Image Restoration Problem

Abstract

In this paper, the problem of color image restoration using a neural network learning approach is addressed. Instead of explicitly specifying the local regularization parameter values, we modify the neural network weights, which are considered as the regularization parameters. These are modified through the supply of appropriate training examples. The desired response of the network is in the form of estimated value for the current pixel. This estimate is used to modify the network weights such that the restored value produced by the network for a pixel is closer to this desired response. In this way, once the neural network is trained, images can be restored without having prior information about the model of noise/blurring with which the image is corrupted.