基于NSST多尺度熵的红外与可见光图像融合  

Infrared and visible image fusion based on NSST multi-scale entropy

作　　者：任亚飞[1] 张娟梅[1] REN Yafei;ZHANG Juanmei(Luoyang Institute of Science and Technology,Luoyang 471023,China)

机构地区：[1]洛阳理工学院,河南洛阳471023

出　　处：《兵器装备工程学报》2022年第7期278-285,共8页Journal of Ordnance Equipment Engineering

基　　金：校级教育教学研究重点项目(2020JYZD-04);国家级创新训练项目(202111070008)。

摘　　要：针对红外图像在恶劣环境依据热辐射成像的显著效果,可见光图像具有较好的三维空间分辨率和图像清晰度,提出基于NSST多尺度熵的红外与可见光图像融合方案。方法区别于传统单一模式的图像融合方案,融合原则基于像素和区域的结合,首先将红外和可见光图像进行NSST多尺度分解,然后在不同的尺度分别对低频和高频系数进行融合,其中融合时权值的选择采用多尺度熵,最后将各个尺度上融合的图像进行逆NSST变换恢复原尺度上的图像。采用两组道路和室外图像进行验证,对比多种不同策略的融合结果,用11种图像指标评价不同的融合质量,给出该次融合的优势和可行性。结果证明采用红外和可见光传感器的图像融合,可以改善单类型传感器的图像采集结果,采用基于NSST多尺度熵的融合方案可以提高融合图像的特征值性能,满足融合图像的多方面指标需求,保持融合算法的可行性、实用性和实时性。Aiming at the remarkable effect of infrared image based on thermal radiation imaging in harsh environment,and the visible image has good three-dimensional spatial resolution and image clarity,an infrared and visible image fusion scheme based on NSST multi-scale entropy is proposed.Method is different from the traditional single mode image fusion scheme.The fusion principle is based on the combination of pixels and regions.First,the infrared and visible images are decomposed by NSST and multi-scale.Then the low-frequency and high-frequency coefficients are fused at different scales.The multi-scale entropy is used to select the weight in fusion.Finally,the fused images on each scale are restored by inverse NSST transform.The experiment uses two groups of road and outdoor images to verify the results,compares the fusion results of different strategies,evaluates different fusion quality with 11 image indexes,and gives the advantages and feasibility of this fusion.The results show that the image fusion of infrared and visible sensors can improve the image acquisition results of single type sensors.The fusion scheme based on NSST multi-scale entropy can improve the eigenvalue performance of the fused image,meet the various index requirements of the fused image,and maintain the feasibility,practicability and real-time of the fusion algorithm.

关 键 词：非下采样剪切波变换 拉普拉斯金字塔 像素级融合 图像融合 多尺度熵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]