一种基于低秩矩阵逼近的图像去噪算法  

An Algorithm for Image Denoising Based on Low Rank Matrix Approximation

作　　者：诸亮[1] 周林强[2] 李一啸[3] 黄朝耿[3] 

机构地区：[1]浙江工业大学浙江省信号处理重点实验室,浙江杭州310023 [2]浙江工业大学信息学院,浙江杭州310023 [3]浙江财经大学信息学院,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2015年第4期36-40,共5页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家自然科学基金资助项目(61273195);浙江省自然科学基金资助项目(61304124;LQ14F030008;61403338)

摘　　要：图像去噪作为图像处理学科的一个重要分支已经被广泛研究许多年,但依然活跃而充满挑战。通过利用图像的非局部自相似性达到去噪目的,是当前较为新颖的去噪方法。结合非局部自相似性和低秩矩阵逼近,提出了一种新的图像去噪算法,并对算法做了分析。仿真结果表明,算法与经典图像去噪算法相比,复杂度更低,去噪效果更好。Image denoising has been extensively studied for decades as an important branch of image processing. It＇s still full of challenges and opportunities. Image denoising by no-local self-similarity has been well acknowledged as an important approach to image denoising. An algorithm based on no-local self-similarity and low rank matrix approximation is proposed. Compared with some classical algorithms,the algorithm this paper proposed has lower complexity and better performance. The experimental results show the superiority of the novel algorithm.

关 键 词：图像去噪 非局部自相似 低秩矩阵逼近 核范数 奇异值分解 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]