A new image registration method for multi-frequency airborne high-resolution SAR images

High resolution airborne synthetic aperture radar (SAR) is useful for surveillance and remote sensing applications. The higher resolution leads to more speckle ad well as a need for more accurate pixel-level registration. Also, the curvature of the Earth needs not to be taken into account while registering two images because the swath of terrain being imaged in an airborne SAR is small. This article describes a new technique for registering multi-frequency airborne high-resolution SAR images. The image registration is set forth based on the fundamental principle of relaxation method, which uses image segment comprising of the relaxation registration results of point feature as global control. In order to guarantee the accuracy and reliability of image registration, line moment of edge feature is first used as match units to develop the traditional relaxation method. Airborne SAR images in Ku-band and X-band are used to test the new image registration method. Experimental results show that the proposed method outperforms standard image registration techniques in terms of reliability and rapidity in most cases. It also achieves better performance than wavelet pyramid method.