Image registration is a fundamental task in image processing, which geometrically matches two or more images taken at different time, from different sensors or from different viewpoints.It is a process of aligning two images into a common coordinate system to monitor and analyze the subtle changes between the two.The image registration methods are mainly divided into two types: spatial domain registration method and frequency domainregistration method.For the processing of series remote sensing images, two representative algorithms of each type were implemented by making the purification to get rid of outliers and Fourier transform in certain neighborhaod according to the actual situatton, and were compared.The experiment results show that the frequency domain method can achieve more accurate registration.At the end of the paper, the spatial domain algorithm is analysed with respect to the experiment results, and a conclusion is offered.%图像配准是图像处理的基本任务之一,用于将不同时间、不同传感器、不同视角及不同拍摄条件下获取的两幅或多幅图像进行(主要是几何意义上的)匹配.它是一个把两幅图像时齐到同一个坐标系下以分析其间细微变换的处理过程.实现图像配准主要有空间域方法和频率域方法两类.针对对序列遥感影像的处理,实现了这两类方法中各具代表性的算法,并在这个过程中根据实际情况做出剔除外点的提纯决策和特定倍数的区域傅里叶变换,然后进行比较.实验表明,频率域方法能取得更精确的配准结果.最后,结合实验结果对空间域算法进行了分析,得出了结论.
展开▼
