The speed and the precision are the two most important criteria for the result of image registration. Scale Invariant Feature Transform (SIFT) is a classical algorithm for feature extraction and description. The SIFT has the characteristics of scale invariance and rotation invariance, while the 128D descriptor has a big adverse effect on the speed of image registration. To improve the time-consuming descriptor, we propose a novel method. The SIFT feature points are extracted at first. Then, ORB algorithm is used to define the main orientation of the points. Based on the moments features of the feature points'neighborhood image block, a 8D descriptor is formed for describing the SIFT feature points of the image and realize SAR image registration. The experimental results demonstrate that the proposed method can improve the registration speed greatly while keeping the image registration precision.