Abstract: The emerging compressed sensing (CS) theory，which includes the incoherent measurement matrix，sparse representation，and nonlinear signal reconstruction，has been employed in the magnetic resonance imaging (MRI).This paper focuses on Split Bregman algorithm which transforms the problem of convex relaxation to the problem of convex optimization.The main advantages of Split Bregman lie in its high computational efficiency and its capacity to solve multi-regularized inverse problem with high accuracy in MRI reconstruction.The function and tuning mechanism of regularization parameters are analyzed theoretically firstly，and then the influences of tuning the regularization parameters on the convergence accuracy and speed are investigated.In this way,the guidelines are provided for choosing suitable parameters in practical applications.

Key words: compressed sensing, magnetic resonance imaging, nonlinear inversion, convex relaxation Split-Bregman regularization algorithm