Synthetic boundary conditions for image deblurring (Q2431145): Difference between revisions

The typical inverse problem of image deblurring is addressed in term of conventional Tikhonov regularization [cf. \textit{A. N.Tikhonov} and \textit{V.Ya. Arsenin}, Methods for the solution of ill-posed problems. (Metody resheniya nekorrektnykh zadach). 2nd ed., rev. and suppl. (Russian) Moskva: ``Nauka''. 286 p. (1979; Zbl 0499.65030)]. The so-called synthetic boundary conditions are introduced to continue important image structures like edge directions and texture outside the viewable region. The linear algebraic structure of the new boundary condition allows for efficient implementation of iterative image deblurring algorithms, and construction of effective preconditioners. The extensive numerical experiments are included. In particular the following practical cases are addressed: Gaussian blur, diagonal motion blur, Gaussian blur with additive Gaussian noise, diagonal motion blur with additive Gaussian noise, the separate chapter deals with usage of the DCT based preconditioning for convergence acceleration (on example of Gaussian blur). The authors provide the links sharing the \texttt{MATLAB} codes of implementation.