计算机工程

• 图形图像处理 • 上一篇    下一篇

基于分数阶变换和改进最小生成树的图像配准算法

韩毅 1,2,赵凯 1,周晏 3   

  1. (1.安阳工学院 计算机科学与信息工程学院,河南 安阳 455000; 2.华中科技大学 国家数控系统工程技术研究中心,武汉 430000; 3.郑州大学 信息工程学院,郑州 450001)
  • 收稿日期:2016-08-08 出版日期:2017-09-15 发布日期:2017-09-15
  • 作者简介:韩毅(1980—),男,讲师、博士研究生,主研方向为计算机图形图像技术、智能控制;赵凯,讲师、硕士; 周晏,副教授、博士。
  • 基金项目:
    河南省科技计划项目(142102310188)。

Image Registration Algorithm Based on Fractional Order Transformation and Improved Minimum Spanning Tree

HAN Yi 1,2,ZHAO Kai 1,ZHOU Yan 3   

  1. (1.College of Computer Science and Information Engineering,Anyang Institute of Technology,Anyang,Henan 455000,China; 2.National Nc Engineering Research Center,Huazhong University of Science and Technology,Wuhan 430000,China; 3.School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China)
  • Received:2016-08-08 Online:2017-09-15 Published:2017-09-15

摘要: 为改善图像配准的精度和稳定性,提出一种新的鲁棒图像配准算法。定义分数阶变换,强化图像特征信息,联合分数阶与高斯核函数,将图像信号变换为尺度空间,利用尺度不变特征变换提取图像特征点,通过改进最小生成树建立特征点的结构关系,完成图像特征点匹配,引入随机抽样一致性技术降低误匹配。实验结果表明,与基于Harris角点检测的匹配算法、基于随机k-d树的匹配算法以及块匹配算法相比,该算法具有更高的配准精度与鲁棒性。

关键词: 图像配准, 分数阶变换, 最小生成树, 特征配准, 随机抽样

Abstract: In order to improve the accuracy and stability of image registration,a new robust image registration algorithm is proposed in this paper.The Fractional Order Transformation(FOT) is defined to enhance the characteristic information of the image.And the image signal is transformed into the scale space by combining the fractional order and Gauss kernel function,then the image feature points are extracted by scale-invariant feature transform.The Minimum Spanning Tree(MST) is improved to establish the structure relation of feature points for finishing the feature points matching.The random sample consensus algorithm is introduced to reduce mismatching.Experimental results show that this algorithm has higher registration accuracy and robustness compared with the current image registration algorithm based on Harris corners,k-d trees and blocks.

Key words: image registration, Fractional Order Transformation(FOT), Minimum Spanning Tree(MST), feature registration, random sampling

中图分类号: