A blind image restoration system using higher-order statistics and Radon transform

We propose a complete system of blind image restoration. We have no restrictions about the blurring filter causing degradation to the original image (e.g., zero or linear phase filter). The blurring filter estimation part of this system consists of two main steps: the mapping of the blurred image into the Radon transform domain, and the blind blur identification in this domain. The mapping simplifies the computational complexity associated with our system from a 2-D to a 1-D domain. The blind blur identification is based on a modified optimization method which uses cumulants in the estimation of the filter coefficients. The deconvolution step is based on a least squares optimization method. A singular value decomposition (SVD) technique is used in improving the optimization process in the estimation and deconvolution steps. Finally, the inverse Radon transform is computed to get the estimated restored image