We consider a new inner-outer iterative algorithm for edge recovery in image restoration and reconstruction problems.We propose a sequence of dynamically updated regularized least squares problems where the value of the regularization parameter for each problem is determined on-the-fly for each through a hybrid regularization approach.