一种空域和频域相结合的遥感图像亚像素配准方法

New Remote Sensing Image Sub-pixel Registration of Spatial and Frequency Domain

扩展相位相关配准方法只适用于图像间相对运动小于图像尺寸的50%甚至30%的情况,其配准精度随着图像间运动增大而降低且易受边缘效应和混叠效应的影响.针对这种情况,本文做了如下改进:在使用相位相关法之前使用空间域方法粗配准,根据粗配准结果提取重叠区域形成两幅新图像,再对两幅新图像使用扩展相位相关进行精配准,解决了原方法只适用于图像间相对运动较小且精度随相对运动增大而降低的问题;为减弱边缘效应对配准精度的影响,通过实验评价了不同加窗方法,并给与了最优的建议;为减弱混叠效应对配准精度的影响,通过实验讨论了理想低通滤镜的截止频率的取值方法.实验证明这种方法不仅拓展了相位相关的使用范围而且在很大程度上提高了配准的精度.

For phase correlation registration method is only applicable to the situation that the relative motion between images is less than 50% or even 30% of image size, and the image registration accuracy become lower with the increase of the motion between ima- ges. This paper made the flowing improvement： use coarse spatial domain registration and extract the public area before using phase correlation method;in order to weaken the edge effects, evaluated different windowing approach and gave the optimal recommenda- tions ;discussed how to obtain the cutoff frequency value of low-pass filter for reduce the aliasing effect. Experiments show that this method can expand the use of phase correlation and largely improve the accuracy of registration.