基于视觉显著性的红外与可见光图像融合算法  

Fusion Algorithm of Infrared and Visible Images via Visual Sali ency

作　　者：段朝伟 徐海刚 张英争 郭泽 张贵锋 孔慧君 石瑞涛 吕冰垚 DUAN Chaowei;XU Haigang;ZHANG Yingzheng;GUO Ze;ZHANG Guifeng;KONG Huijun;SHI Ruitao;LV Bingyao(School of Electronic Information Engineering,Henan Institute of Technology,Xinxiang 453003,China;Engineering Technology Research Center of Intelligent Industrial Equipment,Xinxiang 453003,China;School of Electrical Engineering and Automation,Henan Institute of Technology,Xinxiang 453003,China;Henan Lithium Power Source Co.,Ltd,Xinxiang 453000,China)

机构地区：[1]河南工学院电子信息工程学院,河南新乡453003 [2]河南省工业装备智能化工程技术研究中心,河南新乡453003 [3]河南工学院电气工程与自动化学院,河南新乡453003 [4]河南锂动电源有限公司,河南新乡453000

出　　处：《河南工学院学报》2024年第6期26-30,69,共6页Journal of Henan Institute of Technology

基　　金：河南省科技攻关项目(222102210066);河南工学院高层次人才科研启动基金(KQ2422,XJ2023000602);河南工学院科研反哺教学专项课题(XJ2023004602)。

摘　　要：红外和可见光图像成像原理不同,两种模态图像融合结果常常存在对比度失真、细节信息丢失等问题,给机器视觉任务如自动驾驶、战场侦察等带来了很大挑战。提出了一种基于视觉显著性的红外与可见光图像融合算法。首先,利用高斯滤波器对源图像进行图像分解,获取基本层和细节层;其次,构造对比度显著性融合规则来融合基本层;再次,设计结构显著性融合规则来融合细节层;最后,通过图像重构获得最终的融合结果。实验结果验证了该算法融合结果目标显著、细节清晰、轮廓明显,符合人类视觉感受。Due to the difference between the imaging principles of infrared and visible images,the fusion results of the two multi-modality images often have problems such as contrast distortion and loss of detail information,etc.These problems bring great challenges to machine vision tasks such as automatic driving,battlefield surveillance,etc.A fusion algorithm of infrared and visible images via visual saliency is proposed.Firstly,Gaussian filter is used to decompose source images to obtain base layers and detail layers.Secondly,the contrast saliency fusion rule is constructed to merge the basic layers.Thirdly,the structure saliency fusion rule is designed to merge the detail layers.Finally,the final fusion result is obtained via image reconstruction.Experimental results demonstrate that the presented algorithm can achieve obvious targets,rich details and clear outlines,and it is inconsistent with human visual perception.

关 键 词：红外图像 可见光图像 融合算法 视觉显著性 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]