Image restoration is an ill-posed inverse problem, which has introduced the regularization method to suppress over-amplification. In this paper, by explicitly using the a priori knowledge for bi-level images, we propose a new regularization method for bi-level image restoration. Unlike the well known Tikhonov regularization method which eventually results in a linear system of equations, the new regularization method leads to a nonlinear optimization problem. This nonlinear optimization problem is solved by using the global Barzilai and Borwein gradient method. Simulation results show that the proposed regularization method is feasible and effective for bi-level image restoration.