Blind image deconvolution using space-variant neural network approach

A novel space-variant neural network based on an autoregressive moving average process is proposed for blind image deconvolution. An extended cost function motivated by human visual perception is developed simultaneously to identify the blur and to restore the image degraded by space-variant non-causal blur and additive white Gaussian noise. Since the blur affects various regions of the image differently, the image is divided into blocks according to an assigned level of activity. This is shown to result in more effective enhancement of the textured regions while suppressing the noise in smoother backgrounds.