核磁共振图像归一化互相关非局部自适应去噪  

Adaptive non-local denoising of magnetic resonance images based on normalized cross correlation

作　　者：师黎[1] 许晓辉[1] 陈立伟[1] 

机构地区：[1]郑州大学电气工程学院,郑州450001

出　　处：《计算机应用》2014年第12期3609-3613,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(U1304602);河南省重点科技攻关计划项目(122102210102);河南省高校科技创新团队支持计划项目(2012IRTSTHN001)

摘　　要：为了更好地去除核磁共振(MR)图像中莱斯(Rician)分布的噪声,首先提出使用图像局部归一化互相关(NCC)作为几何结构相似性的一个表征,对传统非局部算法中使用灰度计算像素相似性权值的方法进行有效补充;然后,将改进方法分别应用于非局部均值算法和非局部最小线性均方误差估计算法,并根据局部信噪比(SNR)动态自适应地计算非局部算法中待滤波像素自身的加权值或者像素之间相似性阈值,达到对核磁图像自适应降噪的目的。实验结果表明,该算法可以更好地抑制核磁图像中的莱斯噪声,有效保留图像中细节信息,对核磁共振图像进一步的分析研究以及应用于临床诊断等具有非常重要的应用价值。In order to remove the Rician distribution noise in Magnetic Resonance （MR） images sufficiently, the Normalized Cross Correlation （NCC） of local pixel was proposed to characterize the geometric structure similarity, and was combined with the traditional method of using only pixel intensity to determine its similarity weight. Then the improved method was applied to the non-local mean algorithm and Non-local Linear Minimum Mean Square Error （NLMMSE） estimation algorithm respectively. In order to realize adaptive denoising, the weighted value of pixel to be filtered or the similarity threshold in non-local algorithms were computed according to the local Signal-to-Noise Ratio （SNR） dynamically. The experimental results show that the proposed algorithm not only can better inhibit the Rician noise in MR images, but also can effectively preserve image details, so it possesses a better application value in the further analysis research of MR images and clinical diagnosis.

关 键 词：归一化互相关 莱斯噪声 非局部算法 自适应加权 

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]