基于卷积谱与L0正则化的图像盲去模糊  

Blind image deblurring based on convolutional spectral features and L0-regularization

作　　者：刘光宇[1] 袁权[1] 张令威 吴哲夫 刘光灿 LIU Guang-yu;YUAN Quan;ZHANG Ling-wei;WU Zhe-fu;LIU Guang-can(College of Automation,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学自动化学院,江苏南京210044

出　　处：《计算机工程与设计》2020年第6期1658-1662,共5页Computer Engineering and Design

基　　金：国家自然科学基金优秀青年基金项目(61622305);国家自然科学基金青年基金项目(61502238);江苏省自然科学基金杰出青年基金项目(BK20160040)。

摘　　要：为解决因运动模糊导致的图像关键信息丢失问题,提出一种高效盲去模糊算法。鉴于图像的模糊过程会显著降低图像在频域的表达,同时也会显著改变清晰图像的稀疏性,提出基于图像卷积谱和L0正则化的鲁棒去模糊算法,分别对图像与模糊核进行约束。该算法基于分离变量的方法,保证每个子问题都有良好的收敛性,在文本与自然图像去模糊任务中都取得了良好的效果。To solve the problem of image information loss caused by motion blur,an efficient blind deblurring algorithm was proposed.Based on the observed facts that the blurring process of the image significantly reduces the image’s expression in the frequency domain,and also significantly changes the sparsity of the sharp image,a robust delurring algorithm using spectral properties of convolution operators and L0-regularization was proposed to constrain the sharp image and blur kernel.The algorithm was based on the method of separating variables,which ensured that each sub-problem had good convergence ability,and it achieves good results in both text and natural image deblurring tasks.

关 键 词：盲去模糊 清晰图像 模糊核 卷积算子 L0正则化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]