A PolSAR image classification method which combines traditional polarization decomposition and subaperture analysis is proposed. Firstly, a full resolution PolSAR image is decomposed into several subapertures and two kinds of nonstationary targets are detected. Then, eigenvalue decomposition is done on the coherency matrices of full resolution SAR images. The two parameters, entropy H and angle α are obtained. With these two parameters, targets can be classified based on H-α plane. Finally, combining the nonstationary detection results and the H-α classification results, the classification with higher fineness is done. Simulation results show that better classification effects are obtained with the proposed method.