基于区域生长的多源遥感图像配准

摘要

Multi-source remote sensing images are usually captured by different sensors, in different spectra and/or at different times, which makes them diffcult to match. Although a variety of methods have been proposed to solve this problem, most of them are only suitable for particular applications. It is still an open and challenging task to develop more stable and applicable algorithms. This paper presents a novel registration method based on region growing. It firstly utilizes the global matching to find seed points based on the updated scale-invariant features (SIFT), and then uses the seed points to start the region growing process. In the region growing phase, it estimates the transformation between the sensed image and the reference image employing the matching points in the current region, then expands the searching scope to find other matching points at farther places. This process iteratively executes until the searching region covers the whole image. Combined with bilateral matching, the proposed method can find a large number of evenly distributed matching points from only a few initial correct ones. Experiments show that this algorithm can find a greater number of matching points with higher precision than the existing methods for multi-source remote sensing images with significant gray-scale differences. Therefore, the proposed algorithm is more robust and powerful than several state-of-the-art methods.%多源遥感图像由于成像设备、所用光谱、拍摄时间等因素的不同，给配准带来极大的困难。尽管已经提出了多种匹配方法，但已有方法一般只能适用于特定的应用环境，开发出更加稳定和适用的配准算法仍然是一个极具挑战性的研究课题。提出一种基于区域生长的配准方法，首先，提取改进后的尺度不变特征，通过全局匹配确定种子点和种子区域并完成变换模型的初始化；然后，运用迭代区域生长和双向匹配策略，得到整个图像的可靠匹配点，从而实现多源遥感图像之间的配准。实验表明，该方法提取的匹配点的数量和正确率均远高于已有方法，能够对存在严重灰度差异的多源遥感图像实现高精度的配准，充分证明了该方法的鲁棒性和适用性。