Adaptive image restoration using cellular neural networks
Faculty of Computing, Health and Science
School of Engineering and Mathematics
Image restoration is a process that restores a degraded image to its original or near original form. A review of the current restoration techniques reveals that the Cellular Neural Network (CNN) has not been used to its full potential in image restoration. In particular, CNN have only been used to restore images that have been degraded by either blur or noise but not both artifacts together. In this paper we introduce two masks to be used with the CNN for image restoration. These masks not only have the advantage of being much smaller than the circulant matrix used by many restoration techniques, but also they are able to restore images degraded by blur and noise. Experimental results have shown that the CNN has better performance in image restoration than other types of neural networks, especially in restoring images with high noise content. Due to the non-stationary nature of images, treating an image as a stationary process may not lead to optimum restoration. Alternatively, adaptive restoration approach may be a more suitable approach. This paper describes both the basic and adaptive CNN systems for image restoration and compares their performances.