同步辐射纳米CT图像配准方法研究  

Image alignment for synchrotron radiation based X-ray nano-CT

作　　者：苏博 陶芬 李可 杜国浩 张玲 李中亮[1,2,3] 邓彪 谢红兰[3] 肖体乔 Su Bo;Tao Fen;Li Ke;Du Guo-Hao;Zhang Ling;Li Zhong-Liang;Deng Biao;Xie Hong-Lan;Xiao Ti-Qiao(Shanghai Institute of Applied Physics,China Academy of Sciences,Shanghai 201800,China;University of Chinese Academy of Sciences,Beijing 100084,China;Shanghai Synchrotron Radiation Facility/Zhangjiang Lab,Shanghai Advanced Research Institute,Chinese Academy of Sciences,Shanghai 201204,China)

机构地区：[1]中国科学院上海应用物理研究所,上海201800 [2]中国科学院大学,北京100049 [3]中国科学院上海高等研究院上海光源中心,上海201204

出　　处：《物理学报》2021年第16期47-61,共15页Acta Physica Sinica

基　　金：国家重点研发计划(批准号:2017YFA0206004,2017YFA0206002);国家自然科学基金(批准号:11775297,U1932205)资助的课题。

摘　　要：基于同步辐射的X射线纳米成像技术是无损研究物质内部纳米尺度结构的强大工具,本文总结了图像配准技术在纳米CT成像领域的研究和应用,并根据发展阶段进行分类分析.首先,通过统计近年以来图像配准文献的发表情况,分析并预测纳米尺度图像配准的未来研究方向.其次,基于图像经典配准算法理论,详细介绍了图像配准算法在纳米成像领域最有效的前沿应用.最后,介绍了基于深度学习的图像配准方法的前沿研究,并讨论深度学习在纳米分辨图像配准领域的适用性及发展潜能,根据纳米尺度图像数据的特点及各种深度学习网络模型的特性,展望了同步辐射纳米尺度图像配准技术的未来研究方向及挑战.Synchrotron radiation-based X-ray nano-imaging is a powerful tool for non-destructively studying the internal nano-scale structure of matter.Here in this paper,we review the state-of-the-art image alignment technology in the field of nano-resolution imaging,and classify and analyze the technology according to the research stage.First,through the publications of image alignment algorithm,the development direction of future research is analyzed.Then,the most effective image alignment application in the field of nano imaging based on classic image alignment algorithms is summarized.The paper also presents the feature detection operators that are useful for nano-scale image registration selected from recent feature detection research,which has important guiding significance for the specific application and optimization of nano-imaging image registration.Finally,the state-of-the-art image registration method based on deep learning is introduced,the applicability and potential of deep learning in nano-imaging registration technology are discussed,and future research directions and challenges are prospected based on different neural network characteristics.

关 键 词：图像配准 图像对齐 纳米成像 深度学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]