摘要： 针对光学遥感图像的配准，提出了一种快速有效的图像配准方法：首先分割出参考图像和待配准图像中的高亮度小面元，根据小面元自身的几何特征，获得小面元间的模糊匹配，然后进一步利用RANSAC 方法确立小面元间的精确对应关系；提取匹配小面元的中心作为控制点对，解出仿射变换的参数，从而实现图像的自动配准。利用实际的光学遥感图像进行试验，获得了很好的实验结果。

关键词: 图像分割；不变矩；RANSAC；图像匹配；仿射变换

Abstract: Aiming at the registration of optical remote sensing images, the paper proposes a fast and robust image registration method. First image patches with high gray value are segmented; then combines with the region’s moment invariants and RANSAC method, it can obtain an accurate correspondences of the patches. Extracting the centroids of the matching patches as control points, an automatic registration is performed. The authors have an experiment by real spot and quickbird satellite images and good results are obtained

Key words: Image segmentation; Moment invariants; Random sample consensus (RANSAC); Image registration; Affine transformation