电子学报 ›› 2016, Vol. 44 ›› Issue (7): 1694-1701.DOI: 10.3969/j.issn.0372-2112.2016.07.025

• 学术论文 • 上一篇    下一篇

八通道MSVD构造及其在多聚焦图像融合中的应用

刘斌1, 刘维杰2, 罗益辉1, 郭琳1   

  1. 1. 湖北大学计算机与信息工程学院, 湖北武汉 430062;
    2. 武汉大学计算机学院, 湖北武汉 430072
  • 收稿日期:2014-12-30 修回日期:2015-04-18 出版日期:2016-07-25
    • 作者简介:
    • 刘斌 男,1963年生于湖北红安,教授、博士生导师.2006年在华中科技大学图像所获模式识别与智能系统专业博士学位.主要从事图像融合、小波理论与应用、模式识别等的教学和研究工作.E-mail:liubin3318@163.com;刘维杰 男,1991年生于湖北武汉,现博士就读于武汉大学计算机学院.主要研究方向为信息安全、图像处理
    • 基金资助:
    • 国家自然科学基金面上项目 (No.61471160); 湖北省自然科学基金重点项目 (No.2012FFA053)

Construction of Eight Channel Multi-Resolution Singular Value Decomposition of Matrix and Its Application in Multi-Focus Image Fusion

LIU Bin1, LIU Wei-jie2, LUO Yi-hui1, GUO Lin1   

  1. 1. School of Computer and Information Engineering, Hubei University, Wuhan, Hubei 430062, China;
    2. Computer School, Wuhan University, Wuhan, Hubei 430072, China
  • Received:2014-12-30 Revised:2015-04-18 Online:2016-07-25 Published:2016-07-25
    • Supported by:
    • National Natural Science Foundation of China (No.61471160); Natural Science Foundation of Hubei Province,  China (No.2012FFA053)

摘要:

针对经典的SVD在图像处理中的不足,提出了一种八通道多尺度奇异值分解(Multi-resolution Singular Value Decomposition,MSVD)构造方法,并把它应用于多聚焦图像融合中.首先,在经典SVD的基础上,利用矩阵分块的方法,提出了一种八通道多尺度SVD的构造方法.其次,对参加融合的多聚焦图像进行八通道MSVD分解,得到高层低频和各层七个方向的高频,对分解的低频子图像利用数学形态学增强边缘的方法进行融合、高频子图像采用基于区域能量取大的融合规则进行融合,并重构获得融合结果图像.最后,对融合结果进行主客观评价和分析.实验结果表明,该图像融合方法有较好的视觉效果,结果图像有较高的清晰度,边缘细节信息丰富,没有方块效应.从客观数值和图形评价指标看,该方法有较高的清晰度,其清晰度比基于DWT的融合方法、基于LWT的融合方法、基于Curvelet的融合方法、基于Contourlet的融合方法都高.

关键词: 图像融合, 矩阵奇异值分解, 多尺度分析, 多聚焦图像

Abstract:

To improve the defaults of classical SVD in image processing,a construction method of eight channel multi-resolution singular value decomposition of matrix (MSVD) is presented.An image fusion method based on this MSVD is proposed.Firstly,based on the principle of classical SVD and blocking algorithm,a multi-resolution analysis of eight-channel SVD of matrix is constructed.Each image involved in the fusion are decomposed into one approximation and seven detail images with different resolution by the eight channel multi-resolution singular value decomposition.Secondly,combined with reconstruction algorithm of MSVD,the frame of image fusion is given.The different frequency of original images can be shown in multi-resolution form.The low-frequency sub-image is fused by using an edge enhancement method of mathematical morphological gradient.For the seven high-frequency sub-images of each level,the energy of each image patch over 3×3 window in the high-frequency sub-images is computed as activity measurement.The center pixel of the 3×3 window in which the energy is bigger is selected as the new pixel of the fused result images.Finally,the performance of the result image is evaluated using objective numerical and graphics indices.The experimental results show that the proposed method has good visual effect and has no blocking-artifact.When compared with the fusion method based on DWT,LWT,Curvelet and Contourlet,the proposed fusion method has been observed to have higher definition.

Key words: image fusion, singular value decomposition of matrix, multi-resolution analysis, multi-focus image

中图分类号: