中文摘要：

本文基于Aubert-Aujol（AA）模型和变分水平集方法提出一个新的SAR图像分割模型；在反应-扩散框架下，将各项同性扩散算子加入到该模型的水平集演化方程中，并提出一个两步分裂水平集演化算法，该算法不需要周期性地更新水平集函数。通过对合成图像和Envisat SAR图像的分割实验，表明本文提出的算法具有较准确的边缘定位能力和噪声抑制能力。

英文摘要：

A novel SAR images segmentation model based on Aubert-Aujol （AA） model and variational level set method is proposed. Under the reaction-diffusion framework, by adding a diffusion term into the variational level set evolution equation, a two-step splitting method （TSSM） which does not need to update the level set function periodically is put forward. Segmenting experiments to synthetic images and Envisat SAR images indicate that, compa- ring with classic methods, the proposed algorithm has more accurate boundary positioning and better noise suppression capability.