Convergence of fixed point iteration for deblurring and denoising problem (Q2383623)

The deblurring and denoising problem can be modeled by an equation \(f=Ku+r\), where \(K\) is a known linear blur operator, \(u\) is the true image and \(r\) is the noise. The total variation deblurring and denoising problem are based on a variational problem with contraints with \(u\in L^2(\Omega)\). The Euler-Lagrange equation is a nonlinear algebraic system. \textit{C. T. Vogel} and \textit{M. E. Oman} [SIAM J. Sci. Comput. 17, No.~1, 227--238 (1996; Zbl 0847.65083)] proposed a fixed point iteration to solve this system. This iteration method is analyzed in the present paper. Then a fixed point iteration is derived by half-quadratic regularization. Next, it is shown, how a fixed point iteration based on Bregman distance is obtained. The last iteration method described here uses the generalized Weiszfeld method [cf. \textit{H. Voß} and \textit{U. Eckhardt}, Computing 25, 243--251 (1980; Zbl 0439.65051)]. A proof of the global and linear convergence of the fixed point iteration, as well as an estimate for the rate of convergence is given.