In recent years, the atomic norm minimization (ANM) algorithm has become an important tool in the field of DOA estimation. Aimed at the problem that the DANM is only suitable for single snapshot, an improved decoupled atomic norm minimization algorithm is proposed for both single snapshot and multiple snapshots models. Firstly, through changing the primal optimized model structure in the DANM algorithm, the traditional twodimensional ANM algorithm based on vectorization is further decoupled into two onedimensional ANM solution models to make it suitable for multiple snapshots models. Secondly, in order to avoid the highdimensional calculations brought by the large snapshot, the covariance matrix of the received data and the covariance matrix of the transpose of the received data are used as the calculation data of the two 1D ANM solution model for model solving, so that the size of the dimension of ANM model is limited to the number of sensors. Finally, the DOA of each dimension are obtained by multiple signal classification algorithm, and the 2D DOA estimation is obtained by a simple 2D pairing method. The proposed algorithm maintains the estimation performance advantages of ANM algorithms. Compared with the DANM algorithm, this algorithm improves the estimation accuracy and sparse recovery ability. And compared with the dualitybased 2D ANM algorithm, the CPU time is significantly shortened. The algorithm is valid.