A new regularization method for electrical impedance tomography

Abstract

Regularization methods are good ways to address the ill-posed problem of the inverse problem of electrical impedance tomography (EIT). Common regularization methods such as Tikhonov, L1 norm and total variation (TV) regularization methods have their own inescapable problems, such as the over smoothness of reconstructed edges and the unstable solution due to measurement noise. In this paper, a new regularization method was proposed, which utilizes the total differences of conductivities of neighbored pixels (TDN) as penalty term. This method performed strong robustness against the boundary measurement noise, good ability to reserve the sharp edges as well as special capacity to reconstruct the accurate shape of square objects.