An adaptive fractional-order primal-dual image denoising algorithm

Based on the fractional calculus and the duality principle, a novel fractional-order primal-dual model for image denoising is proposed. We theoretically analyze its equivalency with the fractional ROF model, and its structural similarity with the saddle-point minimax problem. A primal-dual algorithm based on resolvent is extended to fractional-order for solving the model. And the regularization parameter is estimated adaptively in each iteration to guarantee the solution satisfy the Morozov discrepancy principle. We will give the convergence condition for the algorithm and show numerically that the proposed fractional-order denoising model yields good visual effects, and the parameter selection strategy is better than some state-of-the-art methods. © 2015 by Binary Information Press