Image denoising based on mixed total variation regularization with decision-making scheme

Abstract

The denosing method based on total variation has achieved a remarkable denoising performance. However, it usually generates some staircase effects. To overcome the defect of total variation, a novel image denoising method based on total variation is proposed for improving image quality. The present research contains two contributions. Firstly, the mixed total variation model is proposed to suppress staircase effects. Secondly, the optimal threshold and the regularization parameter are all achieved by the decision-making scheme rather than experience. The difference is that the regularization parameter is achieved by the generalized cross-validation approach and the optimal threshold is achieved by the estimated standard deviation of noise. Experiments on some synthetic noisy images and the noisy images on TID2008 database demonstrate that our method is superior to state-of-the-art denoising method in terms of visual quality and objective evaluation.