An AtanTV Nonconvex Regularization Model for MRI Reconstruction

Abstract

Recently, the nonconvex regularization model has attracted much research attention in magnetic resonance imaging (MRI). In comparison with the traditional total variation (TV) regularization method, a nonconvex TV regularizer can effectively improve the fitting performance and prevent bias problems. In this work, we utilize the arctangent function as nonconvex TV regularize term (AtanTV) for MRI reconstruction. With appropriate parameters, the AtanTV model can avoid the systematic underestimation characteristic and maintain the objective function still convex. To address the AtanTV model, we propose an efficient alternating direction method of multipliers (ADMM) method to solve it. Experimental results indicate the superior performance (PSNR, RE, SSIM, etc.) of the proposed MRI reconstruction method.