This paper focuses on developing new and fast algorithms for recovering a sparse vector from a small number of measurements in compressive sensing (CS).
Conventional methods like L1 minimization do not work well for coherent systems, so the paper explores the L1-L2 norm as a superior alternative.
The paper derives an analytical solution for the proximal operator of the L1-L2 metric, making fast L1 solvers like forward-backward splitting (FBS) and alternating direction method of multipliers (ADMM) applicable for L1-L2.
The resulting algorithms are shown to be convergent under mild conditions and significantly more efficient than the original implementation of L1-L2 based on a difference-of-convex approach in numerical experiments.