Alternating direction method of multipliers with variable stepsize for partially parallel MR image reconstruction

The alternating direction multiplier method is an effective method to solve partial parallel imaging (PPI) based on the total variation model, but research results show that the inversion matrix is large and ill-conditioned. In view of the limitation that fixed stepsize of alternating direction multiplier method, the paper presents an adaptive alternating direction multiplier method. The new algorithm combines the traditional alternating direction multiplier method and Barzilai-Borwein (BB) method to deal with non-convex total variation regularization. The experimental results show that the new algorithm not only has obtained better segmentation effect, but also has better convergence and stability.