电子学报 ›› 2018, Vol. 46 ›› Issue (3): 680-687.DOI: 10.3969/j.issn.0372-2112.2018.03.025

• 学术论文 • 上一篇    下一篇

基于变分多尺度的红外与可见光图像融合

冯鑫1, 张建华2, 胡开群1, 翟志芬3   

  1. 1. 重庆工商大学机械工程学院制造装备机构设计与控制重庆市重点实验室, 重庆 400067;
    2. 中国农业科学院农业信息研究所, 北京 100081;
    3. 农业部规划设计研究院, 北京 100125
  • 收稿日期:2016-09-29 修回日期:2016-12-09 出版日期:2018-03-25 发布日期:2018-03-25
  • 作者简介:冯鑫,男,1982年生于四川.现为重庆工商大学机械工程学院副教授.主要研究方向为图像处理、信息融合.E-mail:149495263@qq.com;张建华,男,1982年生于重庆市.现为中国农业科学院农业信息研究所副研究员.主要研究方向为图像处理与分析.E-mail:zhangjianhua@caas.cn;胡开群,女,1981年生于重庆,现为重庆工商大学讲师,主要研究方向为现代农业装备与计算机测控、图像处理技术.E-mail:24507585@qq.com;翟治芬,女,1983年生于山西省阳泉市,现为农业部规划设计研究院高级工程师,主要研究方向为农业技术评估.E-mail:zhaizhifen0821@163.com
  • 基金资助:
    国家自然科学基金(No.31501229);重庆市基础科学与前沿技术研究(一般)项目(No.csct2015jcyjA40014);重庆市教委基金(No.KJ1400628);重庆工商大学青年博士基金(No.1352007);重庆工商大学博士启动基金(No.2014-56-07)

The Infrared and Visible Image Fusion Method Based on Variational Multiscale

FENG Xin1, ZHANG Jian-hua2, HU Kai-qun1, ZHAI Zhi-fen3   

  1. 1. College of Mechanical Engineering, Key Laboratory of Manufacturing Equipment Mechanism Design and Control of Chongqing, Chongqing Technology and Business University, Chongqing 400067, China;
    2. Agricultural Information Institute of Chinese Academy of Agricultural Sciences, Beijing 100081, China;
    3. Chinese Academy of Agricultural Engineering, Beijing 100125, China
  • Received:2016-09-29 Revised:2016-12-09 Online:2018-03-25 Published:2018-03-25

摘要: 为解决变换域融合法对强噪声抑制能力差的问题,提出一种基于变分多尺度分解的红外与可见光图像融合方法.首先对待融合图像分别进行变分多尺度分解,获得纹理分量和结构分量.采用引导滤波的方法进行待融合图像的纹理分量融合,在结构分量融合上提出一种以相位一致性、清晰度、亮度综合信息来权衡融合权重的方法,最后将两幅图像融合后的纹理分量和结构分量相加获取最终融合图像.实验结果从主观观察和客观指标看,本文方法在清晰度和细节信息上比离散小波变换(discrete wavelet transform)法、非下采样轮廓波变换(non-subsampled contourlet transform)法、稀疏表示(sparse representation)法、剪切波变换(shearlet transform)法都要高.

关键词: 红外与可见光, 变分多尺度, 引导滤波, 噪声抑制

Abstract: To improve the ability of the classic transform domain fusion methods filter noise,this paper proposed an infrared and visible light image fusion algorithm based on variation multi-scale decomposition method.Firstly,the original infrared image and visible light image were decomposed into structure components and texture components by variation multi-scale decomposition.The guided filtering method was used in the texture components fusion.In the structure components fusion rule,three coefficients including phase consistency,clarity and brightness information were used to measure the weight.Finally,the performance of the result image is evaluated from objective numerical and subjective observation.When compared with the fusion method based on discrete wavelet transform (DWT),non-subsampled contourlet transform(NSCT),sparse representation(SR) and shearlet transform(ST),the proposed fusion method has higher definition and detail information.

Key words: infrared and visible light, variational multiscale, guide filtering, restrain noise

中图分类号: