抄録
A method is described for estimating the parameters of non-uniformly blurred images and restoring them. An original image is represented by a noncausal stochastic image model, which is driven by white noise. The blur process is modeled using a space-variant linear system. The parameters of the blurred image are estimated by maximizing a function that is the product of the local loglikelihood function of every pixel. A restoring filter on sliding window is applied to every pixel. Testing of the proposed method on simulated images and an actual non-uniformly blurred image showed that it can estimate the radius of out-of-focus blur and restore blurred images.
