Blur identification and image restoration using a multilayer neural network

A neural network model combining an adaptive heteroassociative network and an adaptive autoassociative network with a random Gaussian process is proposed to identify the noncausal blur function and to restore the blurred image at the same time. The noisy blurred images are modeled as continuous associative networks, where the autoassociative part determines the image model coefficients and the heteroassociative part determines the blur function of the system. The estimation and restoration are implemented by using an iterative steepest descent algorithm to minimize the error functions of the networks. Experiment results demonstrate that effective identification and restoration can be performed.>