图像域阈值与维纳滤波组合的图像去噪方法  

Image Denoising by Thresholding and Wiener Filtering in Image Domain

作　　者：周凯汀[1] 郑力新[1] 

机构地区：[1]华侨大学信息科学与工程学院,福建厦门361021

出　　处：《华侨大学学报（自然科学版）》2012年第2期157-162,共6页Journal of Huaqiao University(Natural Science)

基　　金：教育部科研基金重点资助项目(207145);福建省高等学校新世纪优秀人才支持计划项目(07FJRC01)

摘　　要：用小波变换方法获得与带噪图像具有相同尺寸的各尺度与方向的图像域子图,并对各细节子图进行阈值化处理;然后,将去噪的各图像域细节子图与低频子图相加得到初级去噪图像;最后,对初级去噪图像执行图像域维纳滤波,进一步去除噪声斑点.讨论图像域阈值参数的估计方法,提出一种与小波域BayesShrink对应的图像域BayesShrink阈值估计方法.实验结果表明:与小波域阈值或者小波域阈值与图像域维纳滤波组合的方法相比,对于非高度细节的图像,除去低噪声细节相对丰富图像的情况外,图像域阈值与维纳滤波组合在去除平坦区大部分噪声的同时,能更好保留边缘与纹理细节,得到更好的图像质量与更高的峰值信噪比.Each subband image of image domain for every scale and orientation with the same size as the noisy image is obtained by using wavelet transform and each detail subband image is thresholded, then each denoised detail subband image and the approximation image are added together to output the first stage denoised image, at last Wiener filter of image domain is applied to the first stage denoised image for further removal of noisy specks. The method of estimating threshold of image domain is discussed, and a method of estimating BayesShrink threshold of image domain which cori-e- sponds to that of wavelet domain is proposed. Experiment results show that, compared to the method of thresholding in wavelet domain or the method of combining thresholding in wavelet domain and Wiener filtering in image domain, for ima- ges which are not highly detailed, exclude the case of image with relatively more details and low noise strength, combination of thesholding in image domain and Wiener filtering keeps edge and texture details better while eliminating most of the noise in smooth regions, it yields superior image quality and higher peak signal to noise ratio.

关 键 词：图像域 图像去噪 阈值估计 贝叶斯收缩 小波变换 维纳滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]