基于自适应搜索窗的非局部均值去噪算法  

Nonlocal Means De-noising Algorithm Based on Adaptive Search Window

作　　者：胡金蓉[1] 黄增喜[1] 王晓明[1] 杜亚军[1] 

机构地区：[1]西华大学数学与计算机学院,四川成都610039

出　　处：《成都大学学报（自然科学版）》2015年第3期255-258,共4页Journal of Chengdu University（Natural Science Edition）

基　　金：国家自然科学基金(61303126);四川省教育厅自然科学基金(15226443);西华大学重点科研基金(Z1222625);四川省网络智能信息处理高校重点实验室课题(szjj2013-022)资助项目

摘　　要：提出了一种采用局部多项式近似—置信区间交叉(Local Polynomial Approximation and Intersection of Confidence Intervals,LPA-ICI)技术的自适应选取搜索窗的非局部均值图像去噪算法.首先采用LPA-ICI寻找当前像素所在的同质区域,然后将该同质区域设定为当前像素的自适应搜索窗.自适应搜索窗内的像素与当前像素在灰度值以及几何结构上均呈现出"同质"性,对当前像素的估计值更接近真实值.定性与定量实验结果表明:相比于形状和大小固定的搜索窗,自适应选取搜索窗的非局部均值去噪算法能取得更好的去噪效果,对图像中边缘和纹理细节信息具有更好的保护能力.In this paper, a nonlocal means de-noising algorithm based on adaptive search window selection method and local polynomial approximation and intersection of confidence intervals was proposed. Specifically, the homogeneous neighborhood of current pixel was calculated by LPA-ICI at first. Then, the adaptive search window was set for the homogeneous neighborhood. The pixels located in adaptive search window and the current pixel were homogeneous on gray value and geometric structure, so the estimate value for the current pixel was much closer to its true value. The qualitative and quantitative experiments all demonstrated that compared with search window with fixed shape and size, the nonlocal means de-noising algorithm based on adaptive search window had better de-noising performance, especially for the better protection of the edges and textures in images.

关 键 词：图像去噪 非局部均值算法 自适应搜索窗 局部多项式近似—置信区间交叉 

分 类 号：TN911.73[电子电信—通信与信息系统]