On Noise Moment and Range Constrained Image Deconvolution

An approach to the deconvolution of blurred images in additive noise is presented. This approach is based on the use of noise moment and range constraints within a Lagrange optimization framework. Two types of noise moment constraints are examined: standard moments and probability weighted moments. In addition, the range constraints are enforced via an auxiliary mapping such that the optimization can be performed in an unconstrained manner. We report results on several deconvolution experiments and compare them against the Wiener filter so as to make clear the benefits and utility of our approach.