New variant of the HSS iteration method for weighted Toeplitz regularized least-squares problems from image restoration

For fast solving weighted Toeplitz least-squares problems from image restoration, Ng and Pan (2014) studied a new Hermitian and skew-Hermitian splitting (NHSS) preconditioner. In this paper, a generalization of the NHSS preconditioner and the corresponding iterative method are presented. Convergence for the new iteration method is studied and optimal choice of the parameters is discussed. Bounds on the eigenvalues and the corresponding eigenvector distributions are proposed. The degree of the minimal polynomial of the preconditioned matrix is obtained. Numerical experiments arising from image restoration are provided, which show the proposed iteration method is effective and confirm our theoretical results are correct.