数学物理学报(英文版) ›› 2022, Vol. 42 ›› Issue (1): 403-419.doi: 10.1007/s10473-022-0122-z

• 论文 • 上一篇    下一篇

HYBRID REGULARIZED CONE-BEAM RECONSTRUCTION FOR AXIALLY SYMMETRIC OBJECT TOMOGRAPHY

李兴娥1, 魏素花1, 许海波1, 陈冲2   

  1. 1. Institute of Applied Physics and Computational Mathematics, Beijing, 100094, China;
    2. LSEC, ICMSEC, Academy of Mathematics and Systems Science, CAS, Beijing, 100190, China
  • 收稿日期:2020-05-17 修回日期:2021-05-24 出版日期:2022-02-25 发布日期:2022-02-24
  • 通讯作者: Chong CHEN,E-mail:chench@lsec.cc.ac.cn E-mail:chench@lsec.cc.ac.cn
  • 作者简介:Xinge LI,E-mail:li_xinge@iapcm.ac.cn;Suhua WEI,E-mail:wei_suhua@iapcm.ac.cn;Haibo XU,E-mail:xu_haibo@iapcm.ac.cn
  • 基金资助:
    X. Li was supported by National Postdoctoral Program for Innovative Talents (BX201700038). S. Wei was supported by NSFC (11571003). H. Xu was supported by NSFC (11675021). C. Chen was supported by Beijing Natural Science Foundation (Z180002).

HYBRID REGULARIZED CONE-BEAM RECONSTRUCTION FOR AXIALLY SYMMETRIC OBJECT TOMOGRAPHY

Xinge LI1, Suhua WEI1, Haibo XU1, Chong CHEN2   

  1. 1. Institute of Applied Physics and Computational Mathematics, Beijing, 100094, China;
    2. LSEC, ICMSEC, Academy of Mathematics and Systems Science, CAS, Beijing, 100190, China
  • Received:2020-05-17 Revised:2021-05-24 Online:2022-02-25 Published:2022-02-24
  • Contact: Chong CHEN,E-mail:chench@lsec.cc.ac.cn E-mail:chench@lsec.cc.ac.cn
  • Supported by:
    X. Li was supported by National Postdoctoral Program for Innovative Talents (BX201700038). S. Wei was supported by NSFC (11571003). H. Xu was supported by NSFC (11675021). C. Chen was supported by Beijing Natural Science Foundation (Z180002).

摘要: In this paper, we consider 3D tomographic reconstruction for axially symmetric objects from a single radiograph formed by cone-beam X-rays. All contemporary density reconstruction methods in high-energy X-ray radiography are based on the assumption that the cone beam can be treated as fan beams located at parallel planes perpendicular to the symmetric axis, so that the density of the whole object can be recovered layer by layer. Considering the relationship between different layers, we undertake the cone-beam global reconstruction to solve the ambiguity effect at the material interfaces of the reconstruction results. In view of the anisotropy of classical discrete total variations, a new discretization of total variation which yields sharp edges and has better isotropy is introduced in our reconstruction model. Furthermore, considering that the object density consists of continually changing parts and jumps, a high-order regularization term is introduced. The final hybrid regularization model is solved using the alternating proximal gradient method, which was recently applied in image processing. Density reconstruction results are presented for simulated radiographs, which shows that the proposed method has led to an improvement in terms of the preservation of edge location.

关键词: high-energy X-ray radiography, cone-beam global reconstruction, inverse problem, total variation, alternating proximal gradient method

Abstract: In this paper, we consider 3D tomographic reconstruction for axially symmetric objects from a single radiograph formed by cone-beam X-rays. All contemporary density reconstruction methods in high-energy X-ray radiography are based on the assumption that the cone beam can be treated as fan beams located at parallel planes perpendicular to the symmetric axis, so that the density of the whole object can be recovered layer by layer. Considering the relationship between different layers, we undertake the cone-beam global reconstruction to solve the ambiguity effect at the material interfaces of the reconstruction results. In view of the anisotropy of classical discrete total variations, a new discretization of total variation which yields sharp edges and has better isotropy is introduced in our reconstruction model. Furthermore, considering that the object density consists of continually changing parts and jumps, a high-order regularization term is introduced. The final hybrid regularization model is solved using the alternating proximal gradient method, which was recently applied in image processing. Density reconstruction results are presented for simulated radiographs, which shows that the proposed method has led to an improvement in terms of the preservation of edge location.

Key words: high-energy X-ray radiography, cone-beam global reconstruction, inverse problem, total variation, alternating proximal gradient method

中图分类号: 

  • 65R32