华南理工大学学报(自然科学版) ›› 2012, Vol. 40 ›› Issue (8): 51-55.

• 计算机科学与技术 • 上一篇    下一篇

基于压缩传感理论的多聚焦图像融合方法

孙季丰 洪博宇   

  1. 华南理工大学 电子与信息学院,广东 广州 510640
  • 收稿日期:2011-08-23 修回日期:2012-05-11 出版日期:2012-08-25 发布日期:2012-07-01
  • 通信作者: 孙季丰(1962-) ,男,教授,博士生导师,主要从事图像、视频信号处理研究. E-mail:ecjfsun@ scut.edu.cn
  • 作者简介:孙季丰(1962-) ,男,教授,博士生导师,主要从事图像、视频信号处理研究.
  • 基金资助:

    广东省自然科学基金资助项目( 9151064101000037)

Multi-Focus Image Fusion Methods Based on Compressive Sensing Theory

Sun Ji-feng  Hong Bo-yu   

  1. School of Electronic and Information Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2011-08-23 Revised:2012-05-11 Online:2012-08-25 Published:2012-07-01
  • Contact: 孙季丰(1962-) ,男,教授,博士生导师,主要从事图像、视频信号处理研究. E-mail:ecjfsun@ scut.edu.cn
  • About author:孙季丰(1962-) ,男,教授,博士生导师,主要从事图像、视频信号处理研究.
  • Supported by:

    广东省自然科学基金资助项目( 9151064101000037)

摘要: 压缩传感( CS) 理论是对稀疏或可压缩的信号进行采集、编解码的新理论. 文中提出了基于CS 理论的多聚焦图像融合方法. 新方法包括被融合图像的压缩传感变换、像素级融合处理和稀疏重构,其中,采用加权法、对应位置分量最大值法和系数判决准则法进行图像融合处理. 对该方法进行图像融合实验,并同基于小波变换和其它基于CS 理论的融合算法进行比较. 结果表明: CS 融合( 加权法) 得到的信息熵大于CS 融合( 最大值法) ,CS 融合( 系数判决准则法) 则介于CS 融合( 加权法) 和CS 融合( 最大值法) 之间; 用CS 理论进行融合所得到的图像的信息熵均高于采用相同方法基于小波进行融合得到的图像的信息熵.

关键词: 图像融合, 压缩传感, 正交匹配追踪算法, 小波变换

Abstract:

Compressive sensing ( CS) theory is a novel theory used for collecting,encoding and decoding sparse or compressible signals. In this paper,a novel multi-focus image fusion method is proposed based on the CS theory. The method include the compressive sensing-based transform,pixel-level fusion and sparse reconstruction of the fused image,and,specially,performs the fusion by means of the weighted method,the maximum value method and the coefficient judgment method. Then,some image fusion experiments are carried out to compare the above-mentioned method with the wavelet transformation-based image fusion method and other compressive sensing fusion methods. The results show that the entropy obtained through the weighted method is the highest while that through the maximum value method is the lowest; and that the entropy of the image fused according to the CS theory is higher than that through the same fusion method in wavelet.

Key words: image fusion, compressive sensing, orthogonal matching pursuit algorithm, wavelet transform