This paper presents a unified study of L1 over L2 sparsity promoting models in coherent dictionaries.
Theoretical analysis is provided on the existence of global solutions for constrained and unconstrained L1/L2 models.
Sparse properties of local minimizers are analyzed to rule out nonlocal-minimizer stationary solutions.
The alternating direction method of multipliers (ADMM) is applied with nonnegative constraint, resulting in significant improvements in computational time and accuracy for sparse recovery.