Share Email Print

Proceedings Paper

Automatic regularization parameter tuning based on CT Image statistics
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Regularization parameter selection is pivotal in optimizing reconstructed images which controls a balance between fidelity and penalty term. Images reconstructed with the optimal regularization parameter will keep the detail preserved and the noise restrained at the same time. In previous work, we have used CT image statistics to select the optimal regularization parameter by calculating the second order derivates of image variance (Soda-curve). But same as L-curve method, it also needs multiple reconstruction in different regularization parameters which will spend plenty of time. In this paper, we dive into the relationship between image statistics changes and regularization parameter during the iteration. Meanwhile, we propose a method based on the empirical regularity found in the iterations to tune the regularization parameter automatically in order to maintain the image quality. Experiments show that the images reconstructed with the regularization parameters tuned by the proposed method have higher image quality as well as less time when compared to L-curve based results.

Paper Details

Date Published: 1 March 2019
PDF: 6 pages
Proc. SPIE 10948, Medical Imaging 2019: Physics of Medical Imaging, 1094857 (1 March 2019); doi: 10.1117/12.2513115
Show Author Affiliations
Jiayu Duan, Xi'an Jiaotong Univ. (China)
Shaohua Zhi, Xi'an Jiaotong Univ. (China)
Jianmei Cai, Xi'an Jiaotong Univ. (China)
Xuanqin Mou, Xi'an Jiaotong Univ. (China)

Published in SPIE Proceedings Vol. 10948:
Medical Imaging 2019: Physics of Medical Imaging
Taly Gilat Schmidt; Guang-Hong Chen; Hilde Bosmans, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?