一种高精度的亚像素级图像配准算法  

A High Precision Sub-pixel Image Registration Algorithm

作　　者：周玫君 赵辽英[1] 厉小润 ZHOU Meijun;ZHAO Liaoying;LI Xiaorun(School of Computer,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China;College of Electrical Engineering,Zhejiang University,Hangzhou Zhejiang 310027,China)

机构地区：[1]杭州电子科技大学计算机学院,浙江杭州310018 [2]浙江大学信息与电子工程系,浙江杭州310027

出　　处：《杭州电子科技大学学报（自然科学版）》2020年第5期67-72,共6页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家自然科学基金资助项目(61671408);教育部联合基金资助项目(6141A02022350)。

摘　　要：针对图像亚像素配准中旋转缩放参数的快速精确估计问题,提出一种新的亚像素级图像配准算法。首先,采用Fourier-Mellin和相位相关法计算初始旋转缩放系数;其次,利用矩阵乘法离散傅里叶变换快速计算n倍上采样相位相关曲面;再次,对其峰值为中心的局部小区域进行曲面拟合计算旋转缩放系数调整值,修正峰值坐标得到任意精度的旋转缩放参数估计;最后,对待配准图进行反变换,利用基于拟合的扩展相位相关法求亚像素平移参数。仿真图像实验结果表明,相比基于拟合的Fourier-Mellin法和快速上采样Fourier-Mellin法,所提算法的配准精度分别提高0.7079像素和0.2732像素,算法运行时间分别增加约0.0416 s和0.0152 s,说明所提算法能在增加较少运行时间的情况下获得更高的配准精度。真实图像实验结果也验证了所提算法的精度优于其他2种算法。Aiming at the problem of fast and accurate estimation of rotation scaling parameters in subpixel image registration,a new sub-pixel image registration algorithm is proposed.Firstly,the initial rotation scaling coefficient is calculated based on Fourier-Mellin transform and phase correlation.Secondly,the n times upper sampling phase correlation surface is rapidly calculated by matrix multiplication discrete Fourier transform.Thirdly,the adjustment value of rotation scaling coefficient is calculated by fitting the local small area with its peak value as the center.Finally,the subpixel translation parameters are obtained by using the extended phase correlation method based on surface fitting.Experimental results show that,compared with the Fourier-Mellin method based on fitting and the fast up-sampling Fourier-Mellin method,the registration accuracy of the proposed algorithm is improved by about 0.7079 pixels and 0.2732 pixels respectively,and the average running time of the algorithm is increased by about 0.0416 s and 0.0152 s respectively,indicating that the proposed algorithm can obtain higher registration accuracy with minimal time cost.The experimental results of real image also verify that the proposed algorithm is slightly better than the other two methods.

关 键 词：图像配准 傅里叶梅林变换 对数极坐标 曲面拟合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]