基于Split Bregman算法的有限角度CT图像重建  

CT Image Reconstruction from Limited Angle Based on Split Bregman Algorithm

作　　者：苟军年[1] 董海鹰 GOU Jun-nian1 , DONG Hai-ying2(1. School of Automation and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China ; 2. School of New Energy and Power Engineering, Lanzhou Jiaotong University,Wuwei 733099, Gansu, China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070 [2]兰州交通大学新能源与动力工程学院,甘肃武威733099

出　　处：《兰州交通大学学报》2018年第3期37-41,57,共6页Journal of Lanzhou Jiaotong University

基　　金：兰州市科技计划项目(2014-2-7)

摘　　要：对于有限角度投影数据,传统算法重建的CT图像存在伪影严重、噪声明显的现象.基于此,推导建立了CT图像重建的离散化模型,分析了有限角度CT图像重建问题的病态性质,并给出了该问题的L1正则约束求解策略,提出了重建模型的Split Bregman求解算法.仿真实验表明,对于150°有限角度的无噪及有噪投影数据,Split Bregman算法可以获得比传统ART、SART算法更好的主观重建效果,其重建图像的客观评价指标RMSE、AMSE、PSNR和Residuals均优于ART、SART算法重建结果的对应指标.虽然Split Bregman算法重建所需时间较长,但其在有限角度CT图像重建的伪影消除和噪声抑制方面具有一定优势.In regard of the data of limited angle projection,there are obvious defects in traditional CT image reconstruction algorithm,such as serious artifacts and weak noise elimination ability.To solve the problems,the discretized CT image reconstruction model was deduced and set up,and the ill-conditioned property of limited angle CT image reconstruction was analyzed.Then the L_1 regular constraint solving strategy of the problem was given,and the Split Bregman algorithm for the reconstruction model was proposed.Simulation experiments show that for the noiseless and noisy projection data of 150° limited view projection,better reconstruction effect can be obtained by the proposed algorithm compared to the traditional ART and SART algorithms,and the four objective evaluation indexes to the reconstructed image,RMSE,AMSE,PSNRand the Residuals,are all superior to the corresponding indexes of the reconstructed results of ART and SART algorithms.Although longer time for reconstruction is needed for Split Bregman algorithm reconstruction,it has obvious advantages in eliminating the artifacts and noise of the reconstructed image.

关 键 词：CT图像重建 有限角度投影 SPLIT Bregman L1正则化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]