利用多尺度区域权重的红外可见光图像融合  

Infrared Visible Image Fusion Using Multi-Scale Region Weights

作　　者：朱锴 赵巨峰[1,2] 崔光茫 ZHU Kai;ZHAO Jufeng;CUI Guangmang(Institute of Carbon Neutrality and New Energy,Hangzhou Dianzi University,Hangzhou 310018;School of Electronics and Information,Hangzhou Dianzi University,Hangzhou 310018)

机构地区：[1]杭州电子科技大学碳中和新能源研究院,杭州310018 [2]杭州电子科技大学电子信息学院,杭州310018

出　　处：《飞控与探测》2023年第2期52-59,共8页Flight Control & Detection

基　　金：国家自然科学基金(61805063);浙江省自然科学基金(LY22F050002)。

摘　　要：针对红外、可见光图像融合,提出了一种利用多尺度区域权重的方法,获得既具有良好视觉效果又能保留和继承特征细节的融合结果。首先,通过改变平滑滤波的参数构建多尺度分解,获得一系列具有不同细节的双波段图层;其次,基于每一图层生成区域权重图,实现双波段图层融合;最后,利用适当的系数对融合后的图层进行合成,获得最终的融合结果。通过对比实验结果的主客观评价分析,证明此方法能够产生最佳视觉效果的融合图像,能保留和突出原始双波段图像中的细节与特征,并能获得最佳的客观评价指标。This paper proposes an infrared and visible image fusion method using the multi-scale region weight.This method can obtain a result with good visual effects and retain and inherit the feature details from original images.Firstly,a multi-scale decomposition is constructed by changing the smoothing filter parameters to obtain a series of dual-band layers with different details.Secondly,the region weight map is generated based on each layer to achieve dual-band layer fusion.Finally,the fusion layers are synthesized using appropriate coefficients to obtain the final fusion result.Based on the experimental results comparison,both the subjective and objective evaluation indicates that the proposed method can produce the best visual effect fused image,retain and highlight the details and features of the original dual-band images,and obtain the best objective evaluation index.

关 键 词：图像融合 红外与可见光 区域权重 多尺度 

分 类 号：TN911.73[电子电信—通信与信息系统] TP751.1[电子电信—信息与通信工程]