In addition to spectral features, the high spatial resolution (abbreviated as high resolution) remote sensing images contain rich texture features. In order to achieve high precision segmentation for high resolution remote sensing images, this paper proposes a segmentation method which combines multi-feature and fuzzy preference relation. Firstly, statistical features are defined by using pixel spectral measure. The feature image segmentation is realized according to the defined features and the segmentation results are used to construct the fuzzy decision matrix. Then, the fuzzy preference relation matrix is defined based on the pixel to calculate different features weight, and the fuzzy decision matrix is weighted to highlight the superior features and suppress the inferior features. Finally, the optimal image segmentation results are obtained by the defuzzification decision matrix. The qualitative and quantitative evaluation of the synthetic and real high resolution remote sensing image segmentation results show that the total accuracy of synthetic image is 99.8%, Kappa value is 0.998. The proposed algorithm combines the advantages of each feature, which obtains highly accurate segmentation results.