%0 Journal Article
%T 基于自相似性和低秩表示的有噪模糊图像盲复原算法<br>Blind deconvolution method based on self-similarity and low-rank representation for noisy deblurring images
%A 王宇桐
%A 禹晶
%A 肖创柏
%J 北京交通大学学报
%D 2018
%R 10.11860/j.issn.1673-0291.2018.05.017
%X 摘要 图像盲解卷积算法指的是当点扩散函数未知时，利用模糊图像复原出原始清晰图像.图像盲复原不仅是一个典型的反问题，还是一个病态问题.本文提出一种基于自相似性和低秩表示的图像盲复原算法，该算法利用图像不同尺寸或相同尺寸间的自相似性，将图像中的非邻域相似图像块组成一个相似图像块组，对该相似图像块组进行整体的低秩表示.实验表明：该算法能够准确地估计模糊核，复原清晰图像，去除噪声，并具有很好的鲁棒性.<br>Abstract：Blind image deblurring aims to recover the latent sharp image from a noise-blind image when the blur kernel is unknown. Image deblurring is a typical ill-posed inverse problem. In this article, an image blind deconvolution method is proposed based on self-similarity and low-rank representation. The self-similarity between different scales is used, and similar patches in image are combined into a group. Then the group is processed as a whole using low-rank representation. Experimental results demonstrate that the blur kernels estimated by the proposed method are accurate, the restored images have high visual quality without noise, and the method is very robust.
%K 信息处理
%K 低秩表示
%K 结构自相似
%K 盲解卷积
%K 去噪
%K 去模糊&lt
%K br&gt
%K information processing
%K low-rank representation
%K self-similarity
%K blind deconvolution
%K denoising
%K deblurring
%U http://jdxb.bjtu.edu.cn/CN/abstract/abstract3389.shtml