Abstract: Image registration is an important step in the process of super-resolution image reconstruc-tion.In practical applications when the rotation angle is greater than 6°,the Keren algorithm will cre-ate greater registration error,and its computational complexity as the amount of image shift will growseveral times larger.Considering the Vandewalle algorithm,its registration under certain rotation an-gle conditions is a certain advantage,but the overall registration accuracy is lower than the Keren al-gorithm.Aiming to solve the problems of two existing registration algorithms,we combined Adjust-ment of Indirect observations with Constraints to proposed an improved image registration method.The proposed algorithm calculates the rotation parameters and translation parameters using theVandewalle algorithm between the images.Then,the rotation parameters are included as the knownvalues used in Keren algorithm,while the translation parameters are taken as observations,Using theadjustment of indirect observations with constraints iterative shift parameter correction values wecame to the final registration parameters.Our studies show that the algorithm avoids errors stemmingfrom the Taylor series expansion of angle when using the Keren algorithm In the case of large angles,improving the registration accuracy with better reconstruction results.