Identification of causal non-minimum phase blurs using output cumulants

We propose a new blur identification method based on cumulants of an observed noisy blurred image. We show that the new method is capable of identifying blurs that are nonminimum phase which are often encountered in real applications. The blurred image is modeled as a non-minimum phase moving average (MA) model whose parameters are identified using cumulants of the observed image. The new method is superior to model based maximum likelihood and least squares methods that can not identify non-minimum phase blurs since they utilize second-order statistics which are phase blind