华南理工大学学报(自然科学版) ›› 2010, Vol. 38 ›› Issue (10): 68-73,78.doi: 10.3969/j.issn.1000-565X.2010.10.013

• 电子、通信与自动控制 • 上一篇    下一篇

基于相位相关和重采样的亚像素图像配准算法

周武1,2  胡跃明1,2   

  1. 1.华南理工大学 精密电子制造装备教育部工程研究中心, 广东 广州 510640; 2.华南理工大学 自动化科学与工程学院, 广东 广州 510640
  • 收稿日期:2010-01-08 修回日期:2010-04-14 出版日期:2010-10-25 发布日期:2010-10-25
  • 通信作者: 周武(1984-),男,博士生,主要从事计算机精密检测研究. E-mail:zhouwu787@126.com
  • 作者简介:周武(1984-),男,博士生,主要从事计算机精密检测研究.
  • 基金资助:

    国家自然科学基金重点资助项目(60835001)

Sub-Pixel Image Registration Algorithm Based on Phase Correlation and Image Resampling

Zhou Wu 1.2  Hu Yue-ming 1.2   

  1. 1.Engineering Research Center for Precision Electronic Manufacturing Equipments of the Ministry of Education,South China University of Technology,Guangzhou 510640,Guangdong,China;2.School of Automation Science and Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2010-01-08 Revised:2010-04-14 Online:2010-10-25 Published:2010-10-25
  • Contact: 周武(1984-),男,博士生,主要从事计算机精密检测研究. E-mail:zhouwu787@126.com
  • About author:周武(1984-),男,博士生,主要从事计算机精密检测研究.
  • Supported by:

    国家自然科学基金重点资助项目(60835001)

摘要: 为了实现高精度的图像配准,提出了一种基于相位相关和重采样的亚像素图像配准算法.首先基于相位相关实现像素级的粗定位,然后在粗定位点邻域范围内利用矩阵乘法的离散傅里叶变换(DFT)高倍数重采样,并基于相位相关作重采样区域的像素级定位,实现亚像素级的细定位.文中从理论上证明了基于矩阵乘法的DFT实现部分区域重采样的方法与基于零填充重采样的方法在计算精度上具有等效性.实验结果表明,文中算法的配准精度、计算效率和抗噪性优于基于交互相关和扩展相位相关的亚像素配准算法.

关键词: 相位相关, 重采样, 亚像素, 配准, 矩阵乘法

Abstract:

Proposed in this paper is a sub-pixel algorithm based on phase correlation and image resampling for high-accuracy image registration.In this algorithm,first,a pixel-level coarse location is realized using the conventional phase correlation method.Then,a fine step is performed,using the matrix multiplication discrete Fourier transform(DFT) to calculate the resampling region around the coarse point and to further locate the resampling region at a pixel level based on the phase correlation.Moreover,the accuracy equivalence between the matrix multiplication DFT and the zero-padding resampling is proved in detail.Experimental results show that the proposed algorithm is superior to the conventional cross-correlation-based and phase-correlation-based sub-pixel registration algorithms in terms of accuracy,efficiency and noise resistance.

Key words: phase correlation, resampling, sub-pixel, registration, matrix multiplication