Iterative data regularization

The practical aspects of data regularization by iterative optimization are discussed in Chapter . Considering two possible formulations of the optimization problem, I show that the data-space formulation (also known as model preconditioning) provides significantly faster convergence and exhibits more appropriate behavior at early iterations than the alternative, model-space formulation. I introduce preconditioning by recursive filtering and apply it in multidimensional problems with the help of Claerbout's helix transform Claerbout (1998a).