An SAR image registration method based on pyramid model
Yuying Zhang, Tao Wu · 2016
SAR images have been used extensively for various targets recognition, targets detection and change detection, and image registration is the critical and basic technology of the above applications, the result of which directly affects the effect of subsequent processing. Traditional image registration methods are mainly designed for optical images which exhibit different characteristics from SAR images due to disparate imaging sensors, so most methods are not suitable for SAR images. In addition, there are many problems in the most existing methods, such as large computational complexity, difficult in automatic registration and engineering implementation. To overcome these shortcomings, a new method for image registration in non-homologous SAR images is proposed in this paper based on the FFT frequency phase coherent method. The proposed method utilizes multi-look technology to obtain different resolution images of reference images and registration images, then takes advantage of the phase coherent to deal with these multi-scale images respectively, getting the registration point. Multi-scale processing can improve the speed of the algorithm. The FFT phase coherent method can ensure the engineering implementation of the algorithm easily, but there are some common problems in this method, such as the matching fuzziness and threshold determination, and the corresponding solution methods are also proposed in this paper. Experiments are conducted over some images in the actual engineering and the results show that the proposed method can obtain registration points effectively and precisely.