Hopfield Network-Based Image Restoration Using Space-Varing Regularization Technique

Image is the main way we get visual information. It's researched that most (about 80%) of the outer information we got come from the message the visual system received. Massive clear image play very important parts in human's social life and scientific research. In the process of the formation, transmission and display of the image, for many reasons, like the faultiness of the imaging system, the affection of the transmission medium, the relative motion between the target and imaging system, the noise surrounding, that almost all the image we got degraded in a measure which we called Image Degradation.Image Restoration attempts to recover the original scene from its degraded version. It's the inverse process of image degradation. We can estimate the process of image degradation by means of some a priori knowledge and build the model of degradation, then we find out the corresponding inverse methods to process the image to obtain the optimal estimate. Image restoration is used broadly in many fields, the restoration algorithm is the focus of image restoration research.In this paper firstly I roundly introduced the basic knowledge of digital image processing and analyzed the property of human visual system and the image assessment. Introducing the property of the point spread function of different kind of models of degradation and introducing some kind of noise. Analyzing and introducing the theory and property of image restoration technique, based on these knowledge I explained the process, the reason, the model of image degradation in details. Also in this paper I give a presentation of the principle, the structure and property of artificial neural network, giving an introduction of continuous Hopfield network in details and its work on solving the optimization problem, which was then applied on research of image restoration. I delved into kinds of classical algorithm in image restoration application, like Wiener filter, median filter, regularization filter, blind deconvolution filter and Lucy-Richardson filter, and experimented with these filters on different kind of blurred image on Matlab.Based on the knowledge of the original regularization technique and the property of human visual system, I present a new adaptive regularization method for image...