A hybrid spectral gradient method for removing salt-and-pepper impulse noise

In this paper, based on a two-phase scheme, a hybrid spectral gradient (HSG) method is proposed for impulse noise removal. The noise candidates are first identified by the adaptive median filter; then these noise candidates are restored by minimizing an edge-preserving regularization functional, which is accomplished by the proposed hybrid spectral gradient method. Moreover, based on a nonmonotone line search, its global convergence can be established. Numerical experiments are given to illustrate the efficiency of HSG method for removing salt-and-pepper impulse noise.