关键词: 图像配准, 亚像素, 运动参数估计, 自适应, 图像融合

Abstract: The performance of some current area-based image registration algorithms declines when image transformation parameters are of both wide range and high precision. Concerning this problem, a parameter-adaptive registration algorithm was proposed based on the natures of image transformation in frequency/space domain, and the estimation steps and fusion method for rotation parameter and shift parameter were designed. A set of simulation experiments were implemented to compare the performance of the proposed algorithm with the Vandewalle's and improved Keren's. Mean square error and standard deviation of square error were used as evaluation indicators for registration precision and parameters adaptation. The two indicators of the proposed algorithm are lower than those of the other two methods, which means the proposed algorithm has adaptive ability in wide range parameters estimation and high accuracy of registration.

Key words: image registration, sub-pixel, motion parameter estimation, adaptive, image fusion