针对合成孔径雷达(SAR)图像的特点,本文提出基于灰度直方图的混合偏移Rayleigh分布假设下的最小误差阈值化分割算法,并与现有的基于Gauss和Poisson分布假设下的最小误差分割算法以及经典的Otsu算法作了比较。实验和Kolmogorov-Smirnov检验结果表明对SAR图像而言,基于Rayleigh假设的算法可以取得更好的分割效果。 According to the characteristics of Synthetic Aperture Radar (SAR) images, a minimum error thresholding algorithm based on gray histogram with mixed bias Rayleigh distribution is proposed. Compared with the existing minimum error segmentation based on Gauss and Poisson distribution The algorithm and the classical Otsu algorithm are compared. Experiments and Kolmogorov-Smirnov test results show that for SAR images, the Rayleigh-based algorithm can achieve better segmentation results.