电子学报 ›› 2012, Vol. 40 ›› Issue (9): 1829-1838.DOI: 10.3969/j.issn.0372-2112.2012.09.020

• 学术论文 • 上一篇    下一篇

基于参数分步估计的红外与可见光图像自动配准算法

廉蔺1, 李国辉1,2, 张军1, 涂丹1   

  1. 1. 国防科学技术大学信息系统与管理学院系统工程系, 湖南长沙 410073;
    2. 国防科学技术大学信息系统工程国防科技重点实验室, 湖南长沙 410073
  • 收稿日期:2011-07-25 修回日期:2011-11-27 出版日期:2012-09-25
    • 作者简介:
    • 廉 蔺 男,1983年生于河南漯河.国防科学技术大学信息系统与管理学院博士研究生.主要研究方向为视觉不变特征提取及描述,多传感器图像自动配准. E-mail:nudtlarry@gmail.com
    • 基金资助:
    • 国家自然科学基金 (No.60902093)

An Automatic Algorithm for Infrared and Visible Image Registration Based on Parameter Step Estimation

LIAN Lin1, LI Guo-hui1,2, ZHANG Jun1, TU Dan1   

  1. 1. Department of System Engineering, School of Information System and Management, National University of Defense Technology, Changsha, Hunan 410073, China;
    2. Science and Technology Foundation on Information System Engineering Laboratory, National University of Defense Technology, Changsha, Hunan 410073, China
  • Received:2011-07-25 Revised:2011-11-27 Online:2012-09-25 Published:2012-09-25
    • Supported by:
    • National Natural Science Foundation of China (No.60902093)

摘要: 提出了一种仿射模型参数分步估计的红外与可见光图像自动配准算法.首先,使用矩阵正交分解方法,将仿射变换的6个自由度分离为易于估计的切变、尺度比例、旋转、尺度缩放以及x和y方向上的平移量等参数;然后基于方向一致性约束和线段间的对齐度分别构建用于参数分步估计的目标函数,并使用SGA(Stud Genetic Algorithm)算法搜索使目标函数取得近似全局最优解的参数值;最后,基于Powell算法对参数估计值进行局部求精.实验结果表明,当两幅需要配准的图像中含有丰富的关联线段及多样的线段方向分布时,本文算法能够利用这些线段间的方向一致性约束和位置分布信息,有效地实现红外与可见光图像的自动配准,且算法具有较好的配准精度.

关键词: 图像配准, 红外图像, 模型分解, 参数估计

Abstract: An automatic registration algorithm between infrared and visible images based on step estimation of affine model parameters is proposed.Firstly,the six degrees of freedom of the affine model are separated into some more easily estimated parameters using matrix orthogonal decomposition method,which are skew,scale ratio,rotation,scaling and translations in x and y directions.Secondly,two objective functions are constructed for step estimation of these parameters based on the orientation consensus constraint and alignment measure between segments,respectively,and parameter values which make objective function approximate the global optimum are obtained using SGA (Stud Genetic Algorithm) algorithm.Finally,the estimated values of the parameters are locally refined using Powell algorithm.The experimental results show that the proposed method can make full use of orientation consensus constraint and location distribution information of segments,and realize automatic registration between infrared and visible images efficiently and precisely on condition that two images to be registered contain abundant corresponding segments and diverse distributions of segment orientations.

Key words: image registration, infrared image, model decomposition, parameter estimation

中图分类号: