摘要
为了更好地利用图像信息和增强图像的视觉效果,图像去噪成为图像处理领域中一个热点问题。针对图像去噪问题,本文在低秩矩阵填充理论的基础上,提出了两种基于加权低秩矩阵填充的图像去噪算法。首先,基于补丁匹配提取相似的补丁组成低秩矩阵;其次,利用相似补丁的性质形成含有缺失项的低秩矩阵;然后,利用加权核范数构建补丁块的去噪模型;最后,基于奇异值阈值分解和优化一最小化奇异值阈值分别求解加权低秩矩阵去噪优化模型,得到基于奇异值阈值分解的加权矩阵填充(SVT-MC)去噪算法和基于优化-最小化奇异值阈值的加权矩阵填充(MMST)去噪算法。实验结果表明,本文所提出的SVT-MC去噪算法和MMST去噪算法对不同程度的混合噪声都具有良好的去噪效果。
In order to make better use of image information and enhance the visual effect of images,image denoising has become a hot issue in the field of image processing.To address the image denoising problem,this paper proposes two image denoising algorithms based on weighted low-rank matrix padding on the basis of low-rank matrix padding theory.Firstly,the low-rank matrix is formed by extracting similar patches based on patch matching;secondly,the low-rank matrix with missing terms is formed by using the nature of similar patches;then,the denoising model of patch blocks is constructed by using the weighted kernel parametrization;finally,the weighted low-rank matrix denoising optimization model is solved based on the singular value threshold decomposition and the optimisation of a minimisation singular value threshold,respectively,to obtain the weighted matrix padding based on singular value threshold decomposition.The SVT-MC denoising algorithm and the MMST denoising algorithm based on the optimization-minimization singular value threshold are obtained.The experimental results show that both the SVT-MC denoising algorithm and the MMST denoising algorithm have good denoising effects on different levels of mixed noise.
作者
范文婧
杨艳
FAN Wenjing;YANG Yan(State Grid Qinghai Sales and Service Center,Xining 810008 Qinghai,China)
出处
《电力大数据》
2021年第3期76-83,共8页
Power Systems and Big Data
关键词
图像去噪
混合噪声
低秩矩阵
加权矩阵填充
去噪算法
image denoising
mixed noise
low-rank matrix
weighted matrix completion
denoising algorithm
