Detecting and Restoring an Out-of-focus Region based on an Inhomogeneous Stochastic Image Model

We investigated images consisting of an out-of-focus region and an in-focus region. The out-of-focus region was detected by applying a fuzzy clustering to the local correlation of the MDF (measurement differencing filter) output. The correlation was strong for the out-of-focus region, and weak for the in-focus region. The parameters of the out-of-focus process were estimated by matching the model correlation of the MDF output to the sample correlation. The detected out-of-focus region was restored by an iterative method. Computer simulation demonstrated that this method is effective.