摘要： 传统Fourier配准算法存在较大插值误差，对于尺度变化大于2的图像配准效果较差。提出一种基于准极坐标的频域图像配准算法。利用准极坐标离散Fourier变换逼近极对数坐标离散Fourier变换，构造等角度的准极坐标网格，当转换为极对数坐标时，仅需要在极径方向上进行线性插值，从而代替极径和极角方向上的双线性插值。实验结果表明，该算法的配准率高于传统Fourier配准算法。

关键词: Fourier变换, 准极坐标, 图像配准, 脉冲函数, 相位相关, 线性插值

Abstract: Traditional Fourier registration algorithm has big interpolation errors, and has bad image registration effect for scale change larger than 2. In order to solve this problem, this paper proposes a frequency domain image registration algorithm based on quasi-polar coordinate. It uses quasi coordinates discrete Fourier transform close to logarithmic coordinates of discrete Fourier transform, establishes the polar coordinate grid of same angle, when it converts into a logarithmic coordinates, only needs to do a linear interpolation on the pole diameter direction, instead of the direction of bilinear interpolation in polar radius and polar angle. Experimental results show that the registration rate of this algorithm is higher than traditional Fourier registration algorithm.

Key words: Fourier transform, quasi-polar coordinate, image registration, impulse function, phase correlation, linear interpolation

中图分类号: 

• TP301.6