The corners of images are rich in image structure information and widely used in image registration. The Harris algorithm is a classical corner extraction algorithm, and Harris corners are invariant to image rotation but sensitive to image scale change, which limits its application in image regis-
tration when the image scale changes. We propose a multi-corner extraction method based on the feature extraction process of the scale invariant feature transform (SIFT) algorithm, which is insensitive to image rotation and scale changes. We use the SIFT descriptor to describe the multi-scale corners to realize image registration. Experiments on optical and SAR images show that compared with the SIFT and Harris algorithm, the proposal reduces the registration time by more than 40% and improves the utilization of the characteristic points by more than one time while maintaining the registration precision.