提出了一种基于图象边缘轮廓信息的多源遥感图象匹配定位方法,其目的是利用定位精度较高的高分辨率遥感图象对低分辨率图象实现子象素级的匹配定位.该方法有效地利用了多源遥感图象中共有的区域结构信息,将特征匹配和最小二乘影象匹配相结合,具有较好的普适性,且运算快速、抗噪性能好.采用该方法进行NOAA AVHRR图象和Landsat TM图象、1:100万数字地图的边缘图象匹配,并应用于NOAA AVHRR图象的几何精校正中,可实现精确的子象素匹配.

A subpixel multisensor image registration method based on edge and contour information is proposed in this paper. The purpose of this method is to realize subpixel matching of low-resolution images by using high-resolution images. By effectively considering the edge and contour information preserved in the multisensor images, and unifying feature-based matching and least-square matching, this method is work well in general. In addition, it is computationally efficient, and robust to noise. The application of image registration between NOAA-AVHRR, Landsat TM and 1∶1 000 000 digital map data have shown that.