基于小波多分辨分析和LSQR的快速EIT图像重建算法  

Fast reconstruction of EIT images based on multi-resolution analysis and LSQR iterative algorithm

作　　者：汪剑鸣[1] 代月霞 王琦[1] 张荣华[2] 王化祥[3] 陈庆良[4] 

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津工业大学电气工程与自动化学院,天津300387 [3]天津大学电气与自动化工程学院,天津300072 [4]天津市胸科医院,天津300000

出　　处：《电子测量与仪器学报》2016年第4期518-525,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家科技支撑计划重点项目(2013BAF06B00);国家自然科学基金(61373104;61402330;61405143);高等学校博士点专项科研基金(20131201120002);天津市应用基础与前沿技术研究计划(15JCQNJC01500);天津市高等学校科技发展基金计划项目(2012ZD03;20140727)

摘　　要：针对LSQR(least square QR-factorization)算法在求解电阻抗层析成像(electrical impedance tomography,EIT)逆问题时,由于矩阵维数高、计算量大而导致重建速度较慢的问题,提出基于小波多分辨分析的LSQR算法(wavelet multi-resolution based least square QR-factorization,WALSQR)。该方法将EIT的图像重建过程投影到低维的尺度空间进行,通过提取有效信息减少数据计算量,明显提高了图像重建速度。同时由于去除了噪声和冗余信息,保证了成像质量。本文将SALSQR方法分别应用于二维、三维EIT成像实验,证明其有效性,并为三维动态EIT图像重建算法的研究奠定了基础。LSQR （least square QR-factorization）algorithm is always used to solve the inverse problem for electrical impedance tomography （EIT）.However,the LSQR method often involves repeatedly solving large-dimensional ma-trix equations,which are computationally expensive.In this paper,the WALSQR （wavelet multi-resolution based least square QR-factorization）algorithm is proposed to improve the real time performance of EIT reconstruction. With the aid of wavelet based multi-resolution analysis,the LSQR solution is projected onto the low-dimension scale space,where important information of reconstruction is contained and noise or artifacts can be omitted.Thus the computational complexity of reconstruction is reduced without affecting the image quality.In order to verity the ef-fectiveness of the new method,the experiments of 2D and 3D EIT imaging are conducted.It lays the foundation for the research of 3D dynamic EIT image reconstruction algorithm.

关 键 词：电阻抗层析成像 图像重建 小波分析 LSQR算法 

分 类 号：TN98[电子电信—信息与通信工程] TP391[自动化与计算机技术—计算机应用技术]