Abstract: Compared with one pass InSAR system with two antennae, the registration of image pair of repeat-pass InSAR system is much more difficult due to parallel tracks are usually not guaranteed. In this paper, a new image co-registration algorithm for repeat-pass InSAR, which combines the Fourier-Mellin transform and correlation-coefficient algorithm, is proposed to perform accurate co-registration for complex image pair having rotation angle due to unparallel tracks. When the method is used as coarse co-registration step, a nonlinear amplitude transformation is introduced as pre-processing to overcome the failure of Fourier-Mellin transformation method in case of image pair with large dynamic amplitude range and non-uniform amplitude distribution, and when used as fine co-registration step, correlation-coefficient algorithm is incorporated to accurately estimate the small rotation angle between image pair. SIR-C/X X-band experimental data are used to test the proposed algorithm through comparing with traditional correlation-coefficient algorithm in both accuracy and run time. The experimental results prove the effectiveness of the proposed algorithm.