The optimal coefficient vector is calculated by maximizing the fusion ratio of polarization parameters of targets to those of clutter in the traditional Generalized Optimization of Polarimetric Contrast Enhancement(GOPCE)，while the distributions of the fusion value corresponding to desired targets and clutter are ignored. A feature selection method and a modified GOPCE model are developed to improve ship detection performance in polarimetric Synthetic Aperture Radar(SAR). A modified optimization criteria is proposed，which considers minimizing variance of targets and clutter as well. Three best polarization features are selected and the optimal coefficient vector is calculated based on the criteria，and the difference between targets and clutter is enhanced in the fused image. Experiment results demonstrate the improvement of the proposed method.