Registration Between Remote Sensing Images Base on Multi-Window Cross-Correlation
基于分窗口相关的遥感图象配准方法

In remote sensing applications, accurate registration is importment for data fusion and detection of object changes. When registering images with substantially different characteristics, the traditional method of tie points give inaccurate results. In this study, a registration method using a multi window cross correlation technique is developed. A moving window with different scales in the target image is cross correlation with a chosen fixed window in the reference image, and the best match is obtained to provide a satisfactory registration by comparing loop algorithm. Using cross correlation technique for separated windows from reference image, all the match location and the ratio between two images can be determined. Three registration tests employed this method were done between TM images at different time, SAR at different time, AVIRIS and aerial photo. Experimental results on sparately acquired TM images, SAR, AVIRIS images and aerial photos show that the described method can produce subpixel accuracy for registraction between complex images, and it is more convenience than the tie points methods.