In order to improve the efficiency of conventional DOA (directional-of-arrival) estimation methods based on the sparse representation of array covariance, an efficient DOA estimation method relying on the direct 2D sparse reconstruction was proposed. The 2D sparse reconstruction model was constructed by using the array covariance matrix. The noise power can be estimated and the influence of noise on DOA estimation can thus be reduced by applying the eigenvalue decomposition. In solving the 2D sparse reconstruction problem, the 2D-SL0 (2D smoothed L0 norm) algorithm was used, which can deal with the 2D data directly, free of matrix vectorization operation. Simulation results show that the efficiency of the proposed method can be improved significantly, and the performance of the proposed method is better than traditional methods under the conditions of low snapshot, low SNR and sparse array sensors, etc.