文献类型：期刊文献

英文题名：DCDFusion: A dual-consistency decoupling infrared and visible image fusion algorithm for low-light scenes

作者：Xie, Wenkuan[1];Pan, Weiguo[1];Zhang, Jiancheng[1];Dai, Songyin[1];Xu, Bingxin[1];Xu, Cheng[1]

第一作者：Xie, Wenkuan

通讯作者：Pan, WG[1];Zhang, JC[1]

机构：[1]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China

第一机构：北京联合大学北京市信息服务工程重点实验室

通讯机构：[1]corresponding author), Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China.|[11417103]北京联合大学北京市信息服务工程重点实验室;[11417]北京联合大学;

年份：2026

卷号：152

外文期刊名：INFRARED PHYSICS & TECHNOLOGY

收录：;WOS:【SCI-EXPANDED(收录号:WOS:001606328700001)】；

基金：This work was supported by the Beijing Natural Science Foundation (4232026, 4242020); National Natural Science Foundation of China (62572057, 62272049, 62171042, 61871039, U24A20331); Academic Research Projects of Beijing Union University (No. ZK10202404); The R&D Program of Beijing Municipal Education Commission (Grant No. KZ202211417048).

语种：英文

外文关键词：Low-light scenes; Infrared and visible image fusion; Low-light adaptive enhancement; Dual-consistency; Object detection

摘要：To generate high-quality images with prominent targets and rich textures in low-light scenes, this paper addresses the issue of high exposure artifacts caused by the weakened linear correlation in existing fusion methods. A dual-consistency decoupling infrared and visible image fusion algorithm (DCDFusion), specifically designed for low-light scenes, is proposed. First, a low-light adaptive enhancement module (LAEM) is developed to adaptively adjust the brightness of the original visible light image, improving its performance under low-light conditions. Second, cross-modal shallow features are extracted using a cross-modal shared feature encoder (CMSFE), and a dual-consistency feature encoder (DCFE) is introduced to capture both global and local image details. Additionally, a dual awareness adaptive fusion module (DAAF) is designed to selectively emphasize or suppress information from different input sources during the fusion process through an adaptive weight allocation mechanism, optimizing feature fusion. Finally, to enhance the fusion quality, Low-light adaptive enhancement loss and dual-consistency decoupling loss are incorporated to improve both correlation consistency and color fidelity in the fused image. Extensive experiments demonstrate the superiority of our algorithm and its ability to improve the overall performance of infrared and visible light target detection on the same benchmark. Its enhanced fusion quality in low-light conditions makes our DCDFusion method suitable for practical applications such as autonomous driving, video surveillance, and nighttime reconnaissance.