针对合成孔径雷达(SAR)图像水体目标的筛选和边缘模糊问题, 根据SAR图像的成像特点, 给出了一种基于区域生长和马尔可夫随机场(MRF)模型的水体目标提取方法。首先, 利用SAR图像中水体表面呈现为均匀的低强度区域特点, 采用分水岭方法进行过分割, 确定具有较大面积的区块作为水体的种子区域; 然后, 利用水体的种子区域统计特征进行区域生长, 得到初始水体目标; 最后, 利用融合观测图像局部均匀性特征的MRF模型对初始分割图像进行精确的目标提取。实验结果显示所提方法能正确地提取SAR图像中的水体目标并精确定位目标边缘。

Aiming at the imaging characteristics of the water bodies of synthetic aperture radar images, and to solve the problem of blurred edges, a water body target extraction method based on region growing and Markov Random Field (MRF) was proposed.Firstly, considering that the water surface in SAR image is a uniform and low-intensity region, we used watershed method for segmentation, and defined the blocks with large area as the seeded region of water body.Then, region growing was carried out using the statistical characteristics of the region, and the initial water body targets were obtained.Finally, the precise water bodies were extracted from the initially segmented images by the MRF model that combines the local uniformity characteristics of observed image.The experimental results show that the proposed method can accurately extract the water body targets in SAR images and accurately locate the target edges.