Active-set methods for basis pursuit

Many imaging and compressed sensing applications seek sparse solutions to large under-determined least-squares problems. The basis pursuit (BP) approach minimizes the 1-norm of the solution, and the BP denoising (BPDN) approach balances it against the least-squares fit. The duals of these problems are conventional linear and quadratic programs. We introduce a modified parameterization of the BPDN problem and explore the effectiveness of active-set methods for solving its dual. Our basic algorithm for the BP dual unifies several existing algorithms and is applicable to large-scale examples.