低秩矩阵在CT图像重建中的应用  

Low-Rank Matrix Technology for CT Image Reconstruction

作　　者：马海英[1] 宣士斌[1,2,3] 向顺灵 

机构地区：[1]广西民族大学信息科学与工程学院,广西南宁530006 [2]广西民族大学广西混杂计算与集成电路设计分析重建实验室,广西南宁530006 [3]广西民族大学中国-东盟研究中心,广西南宁530006

出　　处：《广西民族大学学报（自然科学版）》2016年第3期86-92,共7页Journal of Guangxi Minzu University ：Natural Science Edition

摘　　要：CT图像重建是医学影像学的重要研究课题,但由于噪声对医学CT图像的影响比较大,为了在不牺牲图像精度和空间分辨率的情况下,重建出噪声含量最低的图像,就要选择合适的去噪方法对图像进行预处理.针对于此,笔者提出一种新的CT图像重建算法,重建过程分成两个步骤:首先用低秩矩阵加权核范数最小化(WNNM)进行图像去噪,再用低秩矩阵分解(LRMD)更新CT图像.实验结果表明,提出的方法具有较强的细节保持能力,低秩矩阵的特性简化计算过程,降低算法复杂度,同时保证了重建图像的去噪效果.Computed tomography （CT） image reconstruction is an important research subject in field of medical imaging. But as the heavily influence of the noise in medical CT image, We must choose appropri- ate denoising method for image preproce-ssing to get the lowest noise images, while without sacrificing im- age precision and spatial resolution. To this problem, this paper proposes a new CT image reconstruction al-gorithm, the reconstruction process has two steps： first, the low rank weighted nuclear matrix norm mini- mization（WNNM） which is applied to image denoising. Then a low-rank decomposition of matrix which is used to update CT images. Experimental results show that the proposed method has strong ability to keep the details of the CT images, the characteristics of low-rank matrix to simplify the calculation process, reduce the complexity of the algorithm, and the denoising method has good denoising effect.

关 键 词：低秩矩阵 核范数 CT图像重建 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]