华南理工大学学报(自然科学版) ›› 2017, Vol. 45 ›› Issue (1): 35-41,52.doi: 10.3969/j.issn.1000-565X.2017.01.005

• 电子、通信与自动控制 • 上一篇    下一篇

基于阈值改进Contourlet 变换的图像融合算法

吴鹏1 李雯霖1,2 齐德昱2† 宋文龙1   

  1. 1. 东北林业大学 机电工程学院,黑龙 江哈尔滨 150040; 2. 华南理工大学 计算机科学与工程学院,广东 广州 510640
  • 收稿日期:2016-04-01 修回日期:2016-06-13 出版日期:2017-01-25 发布日期:2016-12-01
  • 通信作者: 齐德昱(1959-) ,男,教授,博士生导师,主要从事计算机系统结构、人机交互、图像处理算法优化研究. E-mail:qideyu@scut.edu.cn
  • 作者简介:吴鹏(1980-) ,男,博士,副教授,主要从事图像处理研究.E-mail:79102570@qqcom
  • 基金资助:

    国家自然科学基金资助项目( 31470714) ; 哈尔滨市科技创新人才研究专项( 2014RFQXJ127) ; 东北林业大学中央高校基本科研业务费专项资金资助项目( 2572014CB14) ; 黑龙江省博士后科研启动基金资助项目( LBH-Q14006)

An Image Fusion Algorithm Based on Threshold-Improved Contourlet Transform

WU Peng1 LI Wen-lin1,2 QI De-yu2 SONG Wen-long1   

  1. 1.College of Mechanical and Electrical Engineering,Northeast Forest University,Harbin 150040,Heilongjiang,China; 2.School of Computer Science and Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2016-04-01 Revised:2016-06-13 Online:2017-01-25 Published:2016-12-01
  • Contact: 齐德昱(1959-) ,男,教授,博士生导师,主要从事计算机系统结构、人机交互、图像处理算法优化研究. E-mail:qideyu@scut.edu.cn
  • About author:吴鹏(1980-) ,男,博士,副教授,主要从事图像处理研究.E-mail:79102570@qqcom
  • Supported by:
    Supported by the National Natural Science Foundation of China( 31470714)

摘要: 为获得到效果更好的融合图像,提出了一种基于结构相似度( SSIM) 阈值自适应 判定融合规则的改进算法.该算法对低频子带采用了基于相关系数离均差的加权求和融 合算法,以保留更多的概貌信息; 对高频子带则先计算待融合图像各个区域的SSIM,然后 取平均值作为阈值,再根据各个区域的SSIM 与阈值的关系自适应地选择高频子带融合 算法,以保留更多的细节信息.实验结果表明,文中改进的图像融合算法可以获得细节更 丰富和边缘更清晰的融合图像,融合图像质量的客观评价指标更优.

关键词: 图像融合, Contourlet 变换, 结构相似度

Abstract:

In order to obtain better fusion images,an improved algorithm on the basis of structural similarity ( SSIM) threshold is proposed to determine fusion rules adaptively.In the algorithm,for low frequency sub-bands,a weighting summation fusion algorithm on the basis of correlation coefficient and deviation from mean is adopted to preserve more general information.For high frequency sub-bands,the SSIM of each region of the image to be fused is calculated first and then the corresponding average value is taken as the threshold.Afterwards,a high frequency subband fusion algorithm is selected adaptively according to the relationship between the SSIM of each region and the threshold,so as to retain more detail information.Experimental results show that the improved image fusion algorithm helps achieve the fused images of richer detail,sharper edge and better objective evaluation index of quality.

Key words: image fusion, Contourlet transform, structural similarity