基于SVD和残差图像的噪声强度估计方法及其在中子图像去噪中的应用  

Noise intensity estimation method based on SVDand residual images for neutron image denoising

作　　者：刘雪 赵辰一 乔双[1] 任德香 潘瑶 LIU Xue;ZHAO Chen-yi;QIAO Shuang;REN De-xiang;PAN Yao(School of Physics,Northeast Normal University,Changchun 130024,China)

机构地区：[1]东北师范大学物理学院,吉林长春130024

出　　处：《东北师大学报（自然科学版）》2024年第3期62-69,共8页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(12105040);吉林省科技厅创新平台(基地)和人才专项青年成长科技计划项目(20210508027RQ).

摘　　要：针对中子图像去噪问题,提出了一种基于奇异值分解和残差图像的方法,实现含噪声图像的噪声强度值估计.通过包含弱纹理的残差图像,获得由噪声引起的奇异值尾部数据,引入双迭代过程,计算含噪图像的噪声强度估计值.根据噪声强度估计值,应用三维块匹配滤波方法(BM3D)有效地去除噪声.实验结果表明:与其他现有的方法相比,该方法能够更加准确地估计模拟图像和中子图像的噪声强度,在保留图像更多纹理细节的同时有效地去除噪声,获得清晰的图像.Aiming at the problem of neutron image denoising,a method based on singular value decomposition and residual images is proposed to estimate the noise intensity of noisy image.Through the residual image containing weak texture,the singular value tail data caused by noise is obtained,and a double iteration process is introduced to calculate the noise intensity estimate of the noisy image.According to the estimated noise intensity,the Block-Matching and 3D filtering(BM3D)is used to remove the noise effectively.The experimental results show that compared with the existing methods,this method can estimate the noise intensity of the simulated image and the neutron image more accurately,and effectively remove the noise while preserving more texture details of the image,and obtain a clear image.

关 键 词：奇异值分解 残差图像 噪声估计 中子图像 BM3D 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]