In order to improve both reconstruction performance and speed for mult iple measurement vectors (MMV) model with arbitrary sparse structure in compressed sensing (CS),we propo se an improved linearized Bregman iteration for MMV (ILBIMMV) algorithm in this paper.Firstly,an MMV mode l with arbitrary sparse structure is given.At the same time,the characteristics of the model are analy zed theoretically.To effectively reconstruct the MMV model with two dimensions (2D),the linearized Bregman itera tion (LBI) is extended.Secondly,the reconstruction speed is improved by accelerating the algorithm′s convergence,which is achieved by optimizing the condition numbers of s ensing matrices.In addition,the preconditioning is used to optimize the condition number.The convergence and th e computational complexity of the proposed algorithm are theoretically analyzed,and the theoretic analysis is proved by the corresponding simulation results.Finally,the simulation results show that the MMV model with arbitrary sparse structure can be accurately recovered by the proposed algorithm.Meanwhile,the proposed algorithm has evident advantages in reconstruction speed.The effectiveness of the ILBIMMV algorithm i s also verified by the inverse synthetic aperture radar (ISAR) imaging results based on real data with differen t signal to noise ratios.