摘要
为了加强对SAR图像乘性相干斑噪声的抑制,并在相干斑抑制的同时有效提升对边缘的保护性能,提出了一种以均值比与变差系数联合构建的非局部平均(NLM)滤波(MR-NLM)算法。首先以搜索窗内各像素与中心像素在相似窗中的局部均值比替代传统高斯加权欧氏距离,构建对乘性相干斑噪声具有恒虚警边缘检测的相似性测量参量;然后以变差系数替代传统常系数,构建对同质区与边缘区具有较好检测能力的自适应衰减因子;最后利用新构建的相似性测量参量和衰减因子联合形成负指数加权系数,实现对SAR图像的非局部加权滤波。实验结果表明:MR-NLM算法与多种传统算法相比,具有更好的抑斑图像视觉效果及相干斑抑制与边缘保护性能,其等效视数提高10%以上,边缘保持指数提高1%以上。
A new non local means(NLM)filtering algorithm(MR-NLM)based on the mean ratio(MR)and the variation coefficient is proposed to reduce multiplicative speckle noise of SAR images and to improve the edge preservation performance effectively.First,the local MR of each pixel to the center pixel in a search window is estimated by the similarity window,and the similarity measure parameter with constant false-alarm edge detection for the multiplicative speckle noise is constructed by using the MR to replace the traditional Gaussian weighted Euclidean distance.Secondly,an adaptive decay factor with good performance in detecting homogeneous and edge regions is constructed by using the variation coefficient to replace the traditional constant coefficient.Finally,negative exponential weighted coefficients are formed by combining the newly constructed similarity measure parameter with the decay factor,and then used to realize the nonlocal weighted filtering for SAR images.Experimental results and comparisons with several traditional filtering algorithms show that the MR-NLM algorithm has better visual effect for despeckled SAR images,better speckle suppression and edge preservation performance,and that the equivalent number of looks increases by more than 10%,and the edge preservation index improves by more than 1%.
作者
朱磊
蔡飞飞
王延年
郭林源
ZHU Lei;CAI Feifei;WANG Yannian;GUO Linyuan(College of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2018年第4期98-104,共7页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(61401347)
陕西省科技厅工业科技攻关资助项目(2016GY-101)
陕西省教育厅自然科学基金资助项目(17JK0343)
西安工程大学博士科研启动基金资助项目(BS1410)
作者简介
朱磊(1979—),男,教授。
