A Fast Method to Approximate Affine-Transform using Control Points with Noise Reduction.

Registration is the £rst challenge in many applications such as. The performance of many applications such as algorithms highly depend on the performance of registration algorithms. In this paper, a robust method to approximate the af£ne transform between two images of a scene is proposed. A general method to approximate the transform is to evaluate the performance of the transform by its square error and use a Least Mean Square Error optimization method to £nd the transform. The proposed algorithm uses the properties of the Fourier transform in complex domain. Our method is more robust in the mismatching error but it also has more computation cost than the LMSE method[l]. Some experimental results have been provided to show the ef£ciency of the proposed algorithm.