With the amount of space debris increasing accordingly with human activities in near space, the threat of space debris to space missions raises the concern of surveillance of these malicious targets. Space target detection based on optical image is a feasible and effective solution for monitoring these malicious targets. However, as the observation platform is nonestationary, image obtained from the telescope need registration for further operation. In this paper, a star image registration algorithm based on joint feature matching is proposed. The star images are firstly denoised by filtering system. Then reference stars are preselected and their features are constructed. By matching the extracted features, a pool of star pairs is established. Transform parameters are derived from the locations of these matched pairs. Experimental results have validated the capability of our algorithm in pixel accurate star image registration. © 2017 SPIE.