Reduction of staircase effect with total generalized variation regularization for electrical impedance tomography

Y Shi, X Zhang, Z Rao, M Wang, Manuchehr Soleimani

Research output: Contribution to journalArticlepeer-review

18 Citations (SciVal)
131 Downloads (Pure)

Abstract

Image reconstruction in electrical impedance tomography is an ill-posed inverse problem. To address this problem, regularization methods such as Tikhonov regularization and total variation regularization have been adopted. However, the image is over-smoothed when reconstructing with the Tikhonov regularization and staircase effect appears in the image when using the total variation regularization. In this paper, the total generalized variation regularization method which combines the first-order and the second-order derivative terms to perform as the regularization term is proposed to cope with the above problems. The weight between the two derivative terms is adjusted by the weighting factors. Chambolle-Pock primal-dual algorithm, an efficient iterative algorithm to handle optimization problem and solve dual problem, is developed. Simulation and experiments are performed to verify the performance of the total generalized variation regularization method against other regularization methods. Besides, the relative error and correlation coefficient are also calculated to estimate the proposed regularization methods quantitatively. The results indicate that the staircase effect is effectively reduced and the sharp edge is well-preserved in the reconstructed image.

Original languageEnglish
Pages (from-to)9850-9858
Number of pages9
JournalIEEE Sensors Journal
Volume19
Issue number21
Early online date2 Jul 2019
DOIs
Publication statusPublished - 1 Nov 2019

Keywords

  • Electrical impedance tomography
  • image reconstruction
  • staircase effect
  • total generalized variation

ASJC Scopus subject areas

  • Instrumentation
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Reduction of staircase effect with total generalized variation regularization for electrical impedance tomography'. Together they form a unique fingerprint.

Cite this