Image Restoration Based on MC Penalty Function

This study addresses the challenge of robust image restoration corrupted by impulse noise. By integrating an L1 norm loss function with a Minimax Concave (MC) penalty function, a novel nonconvex model is proposed. This model effectively mitigates the impact of outliers while enhancing the sparsity of the solution. To tackle the optimization problem associated with nonconvex regularization, the method of multipliers with alternating direction is utilized (ADMM), demonstrating favorable computational efficiency. Experimental results in image restoration indicate that this approach significantly outperforms traditional methods in both quantitative metrics and visual quality.