Journal of Systems Engineering and Electronics ›› 2010, Vol. 32 ›› Issue (1): 183-187.

• 软件、算法与仿真 • 上一篇    下一篇

基于超复数相位相关的彩色图像配准算法

冯巍,胡波   

  1. (复旦大学电子工程系,上海 200433)
  • 出版日期:2010-01-23 发布日期:2010-01-03

Novel color image registration algorithm using quaternion phase-only correlation

FENG Wei, HU Bo   

  1. (Dept. of Electronic Engineering, Fudan Univ., Shanghai 200433, China)
  • Online:2010-01-23 Published:2010-01-03

摘要:

由于数学上的限制,在使用传统相位相关技术对彩色图像进行配准时,必须先将其转化为灰度图像,在此过程中损失了图像的色度信息,导致配准精度的降低。针对此问题,提出了一种具有亚像素精度的彩色图像配准方法。首先通过数学推导得到一种新的超复数相位相关表达式及相关系数与图像位移间的解析表达式,然后使用最小二乘法从相关系数矩阵中,直接估计出图像间存在的位移。该算法能够充分利用彩色图像的灰度和色度信息,提高配准精度。仿真实验结果证明了算法的有效性。

Abstract:

Due to mathematical limitations, traditional phaseonly correlation (POC)-based registration algorithms can only be applied to grayscale images or at most complex images. A color image must be first converted to a grayscale one before performing the POC, during which the chrominance information is wasted. The conventional POC technique is extended to the quaternion field (QPOC), and a subpixel color image registration algorithm based on the QPOC is proposed. The proposed algorithm can not only naturally make full use of the luminance as well as the chrominance information in color images, but also directly estimate the subpixel translational shift from the 2D data array of the QPOC function. Experimental results show the effectiveness of the proposed algorithm.