计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (5): 150-153.

• 图形图像处理 • 上一篇    下一篇

非负矩阵分解和新轮廓波变换的图像融合

王  斐1,梁晓庚1,2,崔彦凯1,武晓军3   

  1. 1.西北工业大学 自动化学院,西安 710072
    2.洛阳光电技术发展中心,河南 洛阳 471009
    3.洛阳电光设备研究所,河南 洛阳 471009
  • 出版日期:2013-03-01 发布日期:2013-03-14

Image fusion combined with NMF and new contourlet transform

WANG Fei1, LIANG Xiaogeng1,2, CUI Yankai1, WU Xiaojun3   

  1. 1.School of Automation, Northwestern Polytechnical University, Xi’an 710072, China
    2.Luoyang Photoelectric Technology Development Center, Luoyang, Henan 471009, China
    3.Luoyang Institute of Electro-Optical Equipment, Luoyang, Henan 471009, China
  • Online:2013-03-01 Published:2013-03-14

摘要: 为提高融合图像保留源图像的信息量和边缘特征,提出了非负矩阵分解和新轮廓波变换的图像融合算法。以具有尖锐频率局部化特征的新轮廓波对循环平移后的源图像进行分解;运用非负矩阵分解实现低通子带融合,采用能量方差测度函数和匹配度函数实现带通子带融合;对各子带信号重构并逆循环平移,得到融合图像。实验结果分析表明,该方法保留了更多的信息量和边缘细节特征,应用效果较好。

关键词: 非负矩阵分解, 尖锐频率局部化, 图像融合, 能量方差测度函数

Abstract: In order to improve the information and edge details of fused image which come from the original images, fusing algorithm of Non-negative Matrix Factorization(NMF) combined new contourlet transform is proposed. The source images are employed with cycle spinning and the coefficients in different scales and directions are obtained by image decomposition using the new contourlet transform with sharp frequency localization. The fusion results of low pass coefficients are obtained by utilizing NMF and the fusion results of band pass coefficients are got by introducing into the energy-variance measure function as well as matching function. The fused image can be acquired through inverse contourlet transform and inverse cycle spinning. Experimental analysis and results show that the fusion algorithm can retain more information and edge details.

Key words: Non-negative Matrix Factorization(NMF), sharp frequency localization;image fusion, energy-variance measure function