Summary

lit this paper, a successful technique that can handle the blur identification issue in multichannel imaging is developed. Using the maximum likelihood (ML) parametric approach, the image is represented as an extended multichannel AR model that includes the cross-channel components as well as noise. Moreover, blur is modeled in the continuous spatial domain to overcome the limitations of the ML technique. It is shown that this formulation allows estimating the optimal blur extent of the two widely used blurs, namely, uniform linear motion or out-of-focus blurs, in noisy multichannel images. Experimental examples demonstrate that superior restoration results are obtained using the identified blur parameters