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 Saliency
段朝伟 1徐海刚 2张英争 1郭泽 2张贵锋 3孔慧君 4石瑞涛 1吕冰垚1
作者信息
- 1. 河南工学院 电子信息工程学院,河南 新乡 453003;河南省工业装备智能化工程技术研究中心,河南 新乡 453003
- 2. 河南工学院 电气工程与自动化学院,河南 新乡 453003
- 3. 河南锂动电源有限公司,河南 新乡 453000
- 4. 河南省工业装备智能化工程技术研究中心,河南 新乡 453003
- 折叠
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.
Abstract
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Key words
infrared image/visible image/fusion algorithm/visual saliency引用本文复制引用
出版年
2024
NETL
NSTL
万方数据