Structure preserving preconditioners for image deblurring (Q2399156)

Iterative regularization methods for image restoration are investigated. The proposed preconditioning technique preserves the structure of the blurring matrix, which is determined by the boundary conditions of the image. The new preconditioning strategy improves both circulant and structured preconditioning for image deblurring. Further the authors present a nonstationary version of structured preconditioning. Numerical tests demonstrate that the new preconditioning methods deliver more accurate image restorations and have higher robustness than the classical BCCB (block circulant with circulant blocks) preconditioning also for very small regularization parameters.