遥感图像配准是图像融合、多光谱分类、环境监测和图像镶嵌等不可缺少的步骤。本文讨论了遥感图像领域中重要的和最新的配准算法, 将配准方法划分为基于区域的配准、基于图像特征的配准、基于混合模型的配准和基于物理模型的配准四类; 描述了四类配准方法中的典型算法, 并分析了它们的优势和不足, 重点概述了基于特征配准中的局部不变特征变换算法。评述了国内外遥感图像配准的发展现状; 指出了遥感图像配准技术中存在的问题, 即多源遥感图像的配准、遥感图像配准的实时性、遥感图像的非线性配准和遥感图像配准的精度评价, 最后展望了遥感图像配准技术的发展前景。

Remote sensing image registration is an indispensable part for remote sensing image fusion, multispectral classification, environmental monitoring, image mosaicing and so on. In this paper, the important and latest registration methods for remote sensing are discussed and are divided into four types, including area-based methods, feature-based methods, hybrid-based model methods and physically-based model methods. Then, the classic algorithms of each type are analyzed respectively, and their advantages and shortcomings are also stated. The scale invariant feature transform algorithms are mainly discussed. Furthermore, the difficulties of remote sensing image registration techniques are summarized, including the multi-source remote sensing image registration, the real-time registration of remote sensing image, the nonlinear registration of remote sensing image and the accuracy evaluation of remote sensing image registration. Finally, the prospects of image registration are pointed out.