CS-MRI Reconstruction Based on the Constrained TGV-Shearlet Scheme

Abstract

This paper proposes a new constrained total generalized variation (TGV)-shearlet model to the compressive sensing magnetic resonance imaging (MRI) reconstruction via the simple parameter estimation scheme. Due to the non-smooth term included in the proposed model, we employ the alternating direction method of multipliers (ADMM) scheme to split the original problem into some easily solvable subproblems in order to use the convenient soft thresholding operator and the fast Fourier transformation (FFT). Since the proposed numerical algorithm belongs to the framework of the classic ADMM, the convergence can be kept. Experimental results demonstrate that the proposed method outperforms the state-of-the-art unconstrained reconstruction methods in removing artifacts and achieves lower reconstruction errors on the tested dataset.