Generalized Krylov subspace methods for l_p-l_q minimization with application to image restoration

Lothar Reichel
Department of Mathematical Sciences
Kent State University
Kent, Ohio, USA

Abstract:

This talk presents new efficient approaches for the solution of l_p-l_q minimization problems based on the application of successive orthogonal projections onto generalized Krylov subspaces of increasing dimension. The subspaces are generated according to the iteratively reweighted least-squares strategy for the approximation of l_p- and l_q-norms or quasi-norms by using weighted l_2-norms. Computed image restoration examples illustrate the performance of the methods discussed.