基于改进NLmeans 算法的显微CT 图像降噪.pdfVIP

2009 年 7 月 北京航空航天大学学报 July 2009 第35 卷第7 期 Journal of Beijing University of Aeronautics and Astronautics Vo l. 35 No.7 基于改进 NL-means 算法的显微 CT 图像降噪 杨民 李俊江 李保磊 (北京航空航天大学机械工程及自动化学院，北京 100191) 摘 要:显微 CT( Computed Tomography) 采用微焦点射线源，射线剂量低，CT 图像噪 声大，对其降噪十分必要.综述了现存主要 CT 图像降噪算法及其优缺点，介绍了 NL( nor山­ cal) -means 算法，根据实验结果分析了其会在图像平滑区域引入人工伪影的不足.根据 NL­ means 算法的不足，在原算法中引入图像的梯度信息，提出了改进的降噪算法，改进算法保持 了原算法优良的降噪功能，并能有效抑制人工伪影，且能够提高图像细节对比度，实验结果验 证了改进算法的有效性. 关键词:显微CT; 降噪;人工伪影;改进 NL-means 算法;对比度 中图分类号: TP 391 文献标识码:A 文章编号: 1001-5965(2009)07-00833-05 Denoising of micro-CT image based on improved NL-means algorithm Li Baolei Yang Min Li Junjiang (School of Mechanical Engineering and Automation , Beijing University of Aeronautics and Astronautics , Beijing 100191 , China) Abstract: Micro-CT( computed tomography) is based on microfocus X-ray source and the ray dose is low. So the noise level of micro-CT image is big and denoising of the CT image is veηnecessary. The current main denoising methods for CT image and their advantages and disadvantages were summaried. The NL( nonlocal)- means algorithm was described. According to the experiment result , the disadvantage of NL-meas algorithm that artificial artifact will be introduced in smooth region of CT image was analyzed. Based on the shortage of the NL-means algorithm , gradient information of image was introduced in the original algorithm and the im- proved algorithm was put forward. The improved algorithm keeps the excellent denioising function and the arti- ficial artifact was restrained effectively. The detail