Abstract: The traditional polarization sensitive array signal processing methods, e.g. parameter estimation and beamformimg, can not achieve excellent performance under the situation of coherent signal sources. An improved decorrelation algorithm is proposed based on polarization smoothing algorithm. By choosing optimal weight vectors for the signal covariance matrixes of each subarray and ensuring the smoothing covariance matrix satisfies the Toeplitz constraint, the correlation between signals is then eliminated. The derivation of optimal weight vector is given and the rank of the smoothing covariance matrix is analyzed. Simulation results verify the effectiveness of the improved algorithm and the algorithm is also suitable for the nonuniform and coherent noise.