A modified L1/2 regularization algorithm for electrical impedance tomography

• Published in: Measurement science & technology Vol. 31; no. 1
• Main Authors: Fan, Wenru, Wang, Chi
• Format: Journal Article
• Language: English
• Published: IOP Publishing 2020
• Subjects: electrical impedance tomography; image reconstruction; regularization
• ISSN: 0957-0233; 1361-6501
• DOI: 10.1088/1361-6501/ab3ed8

A more approximate solution can be obtained when the regularization technique is used to solve the ill-posed and nonlinear problem of electrical impedance tomography (EIT). However, due to the distribution of the ill-posed sensitivity matrix and sharp changes in preserving boundary information, commonly used regularization algorithms have their own limitations, such as filtering out the effective data and reconstructing images with unclear edges. In this paper, a modified algorithm based on L1/2 regularization (MA1/2) was proposed with the L1/2-norm as the form of the penalty term due to its better sparse characteristic. In addition, characteristics of the gradient in eight neighborhoods were utilized to retrieve reasonable data which might be filtered out. Two different dimension operators for the reconstruction quality were adopted accordingly. Both simulation and static experimental results demonstrated that the MA1/2 method significantly decreased the effect of measurement noise and improved the quality of reconstructed images.