Recent Developments in Inverse Problems - Abstract
Morotti, Elena
In this talk I explore and analyse different Total Variation regularization algorithms for the reconstruction of limited angle tomographic images. The linear inverse problem to be solved in this application is characterized by very large dimensions, ill-posedness and lack of information. Iterative methods, as opposed to geometric algorithms such as filtered backprojection, are getting growing interest in the scientific and commercial community, since they can incorporate a priori information on the desired image, at the expense of computational time. Because of ill-posedness, a reliable solution must be computed by regularization methods. Total Variation regularization is widely used for its edge preserving properties, but the practical difficulty is to solve the problem in a short time. In this talk I'll consider and compare some numerical methods for the solution of the Total Variation regularization problem that are not traditionally used in the tomographic reconstruction applications. They are tested on phantoms with a fixed geometric configuration from a real breast tomosynthesis system.