摘要
针对红外与可见光图像融合中存在的热目标信息丢失、边缘结构模糊、细节损失等问题,提出一种多层分解的图像融合算法。首先使用结构纹理分解将源图分解为细节层和结构层,对细节层使用基于结构相似性和L2范数的融合规则融合并增强;然后提出一种结构均值法,将结构层分解为亮度层和基础层,对亮度层使用绝对值取大融合,对基础层设计了一种基于多指标的融合规则进行融合;最后重构各子融合图像得到最终融合图像。为验证算法的有效性,与9种红外与可见光图像融合算法进行对比,使用空间频率、平均梯度、边缘强度、方差、视觉保真度、基于人类视觉感知的指标和信息熵7种客观图像评价指标,在前5种指标上分别取得27.4%、36.5%、38.2%、8.5%和23.5%的提升。实验结果表明,本文算法在有效保留红外热目标的同时较好地保留了边缘结构和纹理细节,且在客观评价指标上取得了更好的效果。
In order to solve the problems of thermal target information loss,edge structure blur and detail loss in infrared and visible image fusion,an infrared and visible image fusion algorithm is proposed based on structure-texture decomposition.Firstly,the source images are decomposed into detail layer and structural layer by structure-texture decomposition,and the detail layer is fused and enhanced by fusion rule based on structural similarity and L2 norm.Then,a structure-average method is proposed to decompose the structural layer into luminance layer and basic layer.The absolute-value-maximum is used to fuse the luminance layer,and a fusion rule based on multi-indicators is designed for the basic layer.Finally,the fused sub-images are reconstructed to get the final fused image.In order to verify the effectiveness of our algorithm,it is compared with nine infrared and visible image fusion algorithms,and seven objective evaluating indicators are used including spatial frequency,average gradient,edge intensity,variance,visual information fidelity,the metric based on human visual perception and information entropy.The first five indicators are improved by 27. 4%,36. 5%,38. 2%,8. 5% and 23. 5%, respectively. The experimental results show that the proposed algorithm not only effectively retains the infrared thermal target, but also retains the edge structure and texture details, and achieves better results in the objective evaluating indicators.
作者
李青松
杨莘
吴谨
黄泽丰
LI Qing-song;YANG Shen;WU Jin;HUANG Ze-feng(School of Information Science and Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)
出处
《液晶与显示》
CAS
CSCD
北大核心
2023年第10期1389-1398,共10页
Chinese Journal of Liquid Crystals and Displays
基金
国家自然科学基金(No.61702384)。
关键词
图像处理
图像融合
结构纹理分解
红外图像
可见光图像
image processing
image fusion
structure-texture decomposition
infrared image
visible image
