基于绝对值光滑化的图像去模糊算法

The Algorithm for Image Deblurring Based on Smoothing Absolute Value Function

结合图像边缘信息及绝对值光滑技术,提出了一个凸的光滑去模糊模型。该模型是在TV和Tikhonov正则化混合模型的基础上,利用光滑函数来近似混合模型的TV正则化项,使其成为光滑凸优化问题,并利用半近端交替方向乘子法(SPADMM)进行求解。在不同的模糊核和高斯噪声水平上的数值实验结果表明,本文算法不仅有效地消除了噪声和模糊,而且较好地保留了图像边缘信息。

A convex smooth deblurring model is presented which combines the edge information of the image and absolute value smoothing technique.Based on the TV and Tikhonov hybrid regularization model,the smoothing function is employed to approximate the TV regularization term of the hybrid model,which makes it a smooth and convex optimization problem,and it is solved by the semi-proximal alternating direction method of multipliers(SPADMM)algorithm.Numerical experimental results on different fuzzy kernels and Gaussian noise levels show that our algorithm can not only effectively eliminate noise and blur,but also retain the image edge information better.