首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于正弦图分区修复的稀疏角度CT重建算法
引用本文:张萌,梁业星,陈燕,桂志国,张权.基于正弦图分区修复的稀疏角度CT重建算法[J].科学技术与工程,2021,21(12):5011-5017.
作者姓名:张萌  梁业星  陈燕  桂志国  张权
作者单位:中北大学电子测试技术国家重点实验室,太原030051;中北大学生物医学成像与影像大数据山西省重点实验室,太原030051;中北大学信息与通信工程学院,太原030051;中北大学电子测试技术国家重点实验室,太原030051;中北大学生物医学成像与影像大数据山西省重点实验室,太原030051;中北大学信息与通信工程学院,太原030051
基金项目:国家自然科学基金(61671413,61801438)、山西省自然科学基金(201901D111153)、电子测试技术重点实验室开放(ZDSYSJ2015006)、中北大学青年学术带头人项目(QX201801)和山西省青年科学(201801D221196)
摘    要:针对稀疏投影CT重建图像中的条形伪影问题,提出一种稀疏表示与低秩矩阵填充相结合的正弦图分区修复方法.首先,将正弦图子块依据灰度熵大小分为两类;然后,采用字典学习算法修复边界区域的正弦图子块,为了保留正弦图的内部结构,设计一种联合修复模型用于内部子块的修复,将正弦图的低秩特性融入稀疏表示模型中,以便引入非局部信息;最后,组成完整的正弦图并经滤波反投影(FBP)重建获得最终图像.实验结果表明,与经典算法相比,该算法在投影域与图像域皆有较优表现,能够较好地修复正弦图的结构,明显改善稀疏重建图像中的条形伪影及结构模糊问题.

关 键 词:稀疏角度投影  字典学习  低秩矩阵  正弦图修复  灰度熵
收稿时间:2020/7/27 0:00:00
修稿时间:2021/4/17 0:00:00

Sparse-view CT reconstruction algorithm based on sinogram divisional inpainting
Zhang Meng,Liang Yexing,Chen Yan,Gui Zhiguo,Zhang Quan.Sparse-view CT reconstruction algorithm based on sinogram divisional inpainting[J].Science Technology and Engineering,2021,21(12):5011-5017.
Authors:Zhang Meng  Liang Yexing  Chen Yan  Gui Zhiguo  Zhang Quan
Institution:North University of China
Abstract:Aiming at the problem of streak artifacts due to sparse projections in CT reconstructed images, a sinogram inpainting method combining sparse representation and low-rank matrix filling was proposed. Firstly, the sinogram sub-blocks were divided into two categories according to their gray entropy. After that the dictionary learning algorithm was used to repair sinogram sub-blocks at the boundary. In order to preserve the internal structure of the sinogram, a combined inpainting model was designed to repair the internal sub-blocks, and the low-rank characteristics of the sinogram were incorporated into the sparse representation model so as to introduce the non-local information. Finally, the repaired sub-blocks were composed into a complete sinogram, and then the final image was reconstructed by the filtered back projection (FBP) algorithm. Experimental results show that compared with the classical algorithms, the proposed algorithm can achieve better performance in both the projection domain and the image domain. It can attain superior sinogram inpainting effect, lead to significant streak artifacts suppression and structure burring improvement of the sparse-view reconstructed image.
Keywords:sparse-view projection  dictionary learning  low-rank matrix  sinogram inpainting  gray entropy
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《科学技术与工程》浏览原始摘要信息
点击此处可从《科学技术与工程》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号