有效保持细节特征的图像椒盐噪声滤除方法  

A Novel Detail-Preserving Algorithm for Removing Salt-and-Pepper Noise

作　　者：肖泉[1] 丁兴号[1] 王守觉[1,2] 廖英豪[1] 郭东辉[1] 

机构地区：[1]厦门大学信息科学与技术学院电子工程系,福建厦门361005 [2]中国科学院半导体研究所,北京100083

出　　处：《电子学报》2010年第10期2273-2278,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.30900328);国家自然科学主任基金(No.60753001);福建省自然科学基金(No.2008J0032;No.2009J01301);厦门大学985二期信息创新平台(No.000-X07204);厦门市科技计划高校创新(No.3502Z20083006)

摘　　要：论文结合几种去噪方法,提出一种统一的图像去噪模型.该模型通过一个统一的目标函数将图像去噪问题转化为最优化问题,目标函数的构造主要包括估计残差惩罚函数、局部权函数及正则化项三个方面.随后基于此模型提出一种新的去除椒盐噪声的非线性滤波方法,其中估计残差惩罚函数采用L1范数形式,局部权函数采用自适应高斯核函数,正则化项则利用图像的小波域稀疏性作为先验约束来构造.由于充分融合了图像的全局和局部统计特性,因而在抑制噪声的同时能够更好地保持图像边缘等细节特征,相关去噪实验结果证实了本文方法的有效性.A unified model for image denoising is proposed based on the analysis of several denoising methods.It by designing a unified objective function leads the image denoising into an optimization problem.The objective function builds up with three parts：the residual error penalty function,the local weighting function and the regularization term.Under this unified model,a novel nonlinear filtering algorithm for removing salt-and-pepper noise is subsequently proposed,in which the residual error penalty function in the form of L1 norm,the weighting function utilizes the adaptive Gaussian kernel,and the regularization term is constructed by using the apriority of image＇s sparse property in wavelet domain.Since the proposed model fully considers the global and local statistical property of image,and hence behaves much better in edge preservation.Experimental results demonstrate the effectiveness of the proposed method.

关 键 词：图像去噪 非线性滤波 自适应加权 细节保持 

分 类 号：TN911.73[电子电信—通信与信息系统]