A new affine transformation parameters estimation method

The precise estimation of affine transformation parameters is most significant in image matching. Since the traditional parameters estimation method based on solving over determined equation is sensitive to mismatch points, high precision of matching is required. In view of this situation, the nonlinear affine transformation parameters estimation method is studied in this paper, the relevant mathematical model is established, and the fast iterative formulas are deduced. Besides that, by combining the parameter estimation method with traditional SIFT feature extraction algorithm, a new image registration algorithm is proposed. The simulation results show that this estimation algorithm is robust to mismatch points, the estimated parameters are more accurate than traditional method, and the time consuming is also decreased. With the improved method, precise registration for actual aviation images are realized, therefore this method is precise and suitable in practical engineering.