提出了一种基于局部相位相关的高效和鲁棒的亚像素级图像配准方法。通过传统的相位相关算法估计出初始平移参数后，在初始位置的引导下对互相关功率谱进行上采样矩阵Fourier变换，实现了图像局部相位相关，得到图像间亚像素级平移参数。实验结果表明，算法配准精度较高，且对随机噪声和光照变化具有较强的鲁棒性。

An efficient and robust subpixel image registration method based on Local Phase Correlation (LPC) is developed in this paper. After estimating the initial translation by the traditional Phase Correlation (PC) method, the initial estimation is refined by the upsampling matrix Fourier transform of cross-correlation power spectrum. Then the translation is tuned by the refinement to achieve subpixel registration. Experiments on various image pairs show that the LPC method achieves higher registration precision, and immunes to random noise and illumination variation.