图像复原的Bregman迭代双正则化方法

易丽娅; 鲁晓磊; 王进军; 王芙蓉

doi:10.11834/jig.20110307

图像处理和编码 | 浏览量 : 0 下载量: 284 CSCD: 0

PDF
导出
分享
收藏
专辑

图像复原的Bregman迭代双正则化方法
Image restoration based on Bregman iterative double regularization
2011年16卷第3期页码：350-356
网络出版：2011-04-01，

纸质出版：2011
DOI： 10.11834/jig.20110307
稿件说明：

移动端阅览

易丽娅, 鲁晓磊, 王进军, 王芙蓉. 图像复原的Bregman迭代双正则化方法[J]. 中国图象图形学报, 2011,16(3):350-356. DOI： 10.11834/jig.20110307.

Image restoration based on Bregman iterative double regularization[J]. Journal of Image and Graphics, 2011, 16(3): 350-356. DOI： 10.11834/jig.20110307.

摘要

正则化图像复原最终会导致一个大规模优化问题，提出了一种基于Bregman迭代双正则化的图像复原方法。该方法中目标函数同时考虑总变分正则化和小波域稀疏正则化，在Bregman框架下解决图像复原问题，并且给出了用于解该问题的分裂Bregman迭代算法。该算法将复杂的优化问题转化为几十次简单的迭代加以解决，每次迭代只需几次快速傅里叶变换和收缩操作即可。实验结果表明，提出的复原算法不论从客观改善信噪比还是主观视觉，都能取得很好的效果。同时与目前的复原算法相比，该算法有更快的收敛速度。

Abstract

To handle the large-scale optimization problem caused by regular image restoration

this paper introduces a novel image restoration method based on Bregman iterative double regularization. In this method

the designed objective function considers both the total variation regularization and the wavelet domain sparsity constraint

and solves the problem under Bregman framework with the split Bregman iterative algorithm. The algorithm converts the complex optimization problem to several iterations

each of which requires only several simple Fast Fourier Transformations and shrinkage operations. The experimental results show that the proposed method improves both the objective SNR and the subjective perceptual image quality with a faster convergence rate compared to existing approaches.