Kaczmarz methods for regularizing nonlinear ill-posed equations. II: Applications (Q2466972)

In part I [ibid. 1, No. 2, 289--298 (2007; Zbl 1123.65051)], the authors analyzed iterative regularization methods of Landweber-Kaczmarz type for solving ill-posed operator equations in Hilbert space. Two cases of such algorithms have been investigated: the looping Landweber-Kaczmarz (lLK) method and the embedded Landweber-Kaczmarz (eLK) method. The stability and the convergence for both the eLK and lLK methods have been proved. In the present paper the authors apply these methods to three different problems: a linear inverse problem related to thermoacustic tomography, an nonlinear inverse problem for semiconductors, and a nonlinear problem in Schlieren tomography. The numerical results are compared to the classical Landweber-Kaczmarz method. For each of the three problems, first, the mathematical model is constructed, then the abstract formulation in Hilbert spaces is deduced, and third, numerical reconstruction is performed. The presented algorithms show robustness, stability, computational efficiency and high accuracy.