Abstract: In the symmetrical distribution ocean noise field, in order to improve the direction-of-arrival (DOA) estima-tion performance in low signal-to-noise (SNR) conditions, a novel MUSIC algorithm that reconstructs the signal cova-riance matrix is presented. Based on the properties that the image part of data covariance matrix has no relation with the symmetric noise, the signal covariance matrix is reconstructed by using the image part of data covariance matrix and the DOA which is estimated by Image-MUSIC algorithm. Thus, the MUSIC algorithm is implemented. Simulation results show that the new algorithm effectively reduces the influence of symmetrical noise on DOA estimation and avoids bi-lateral spectrum arising, moreover the algorithm achieves higher resolution and lower resolution threshold in low SNR cases. Covariance matrix's Toeplitz modification that aims to improve the DOA estimation performance of MUSIC algorithm has also been studied, and simulation results show that the resolution of MUSIC algorithm is improved sig-nificantly.