A Blind Image Restoration Algorithm Based on Cumulants
Hongzhi Wang, Jinming Huo, Chunpeng Song · 2006
A new system of image restoration without requiring the knowledge of blurs parameters based on cumulants is investigated. Blind image restoration is a classical problem that is known to be difficult and have attracted major research efforts. At first, we project the blurred image (a 2-D signal) into a number of one-dimensional signals using Radon transform, and the blind blur identification in this domain. So the difficulties associated with the combination of 2-D image and their cumulants are reduced by means of the Radon transform. Second, the blind blur identification is based on cumulants in the estimation of the filter coefficients; because of using cumulants, the additive noise can be Gaussian colored one. Finally, we choose an iterative deconvolution called Richard-Lucy method (RL) and the inverse Radon transform is computed to get the estimated image. Simulation results illustrate the performance of the proposed method