In order to meet the real-time requirements of image registration,the image registration method based on the curvature scale space algorithm is proposed.Firstly of all,the image corner features are extracted by the curvature scale space algorithm.Then,the improved speeded up robust features algorithm is used to generate the 64-dimensional corner feature descriptor vectors and reduce the descriptor vectors to 24-dimensional.Finally,the improved similarity distance algorithm and the random sample consensus algorithm are used to match.The first simulation experiment shows that the image registration accuracy is equivalent to the traditional scale invariant feature transform algorithm,the traditional speeded up robust features algorithm,and other improved image registration methods,but it has certain advantages in real-time of image registration.The second simulation experiment verified the universal effectiveness of the method through the stereo matching library.