A novel fusion-based ship detection method from polarimetric Synthetic Aperture Radar (Pol-SAR) images is proposed in this paper. After feature extraction and constant false alarm rate (CFAR) detection, the detection results of HH channel, diplane scattering by Pauli decomposition and helical factor by Barnes decomposition are fused together. The confirmed targets and potential target pixels can be obtained after the fusion process. Using the difference degree of the target, potential target pixels can be classified. The fusion-based ship detection method works accurately by utilizing three different features comprehensively. The result of applying the technique to measured Airborne Synthetic Radar (AIRSAR) data shows that the novel detection method can achieve better performance in both ship’s detection and ship’s shape preservation compared to the result of K-means clustering method and the Notch Filter method.