Tie-point extraction is the key step of coregistration of multitemporal images. In this paper, Scale invariant feature transformation (SIFT) is applied to extract tie-points from multitemporal SAR images. A histogram-based preprocessing method and the optimization of SIFT parameters are proposed for increasing the number of correct tie-points. We explore the performance of SIFT algorithm on different topographic ENVISAT SAR images. The experimental results show that these methods are robust to speckle noise and can markedly improve the performance of SIFT on SAR images.