计算机技术与发展2025，Vol.35Issue(11)：69-77,9.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0164
基于复合注意融合策略的多源图像融合算法研究
Research on Multi-source Image Fusion Algorithm Based on Composite Attention Fusion Strategy
摘要
Abstract
Multi-source image fusion is a key research branch in the field of multi-source fusion,and is committed to achieving comple-mentarity and enhancement of data by integrating image information from different modes.This technology plays an important role in many fields,including medicine,industry,safety,fire protection,autonomous driving and consumer electronics,etc.,providing richer and more accurate image analysis and application support for these fields.A new composite attention fusion strategy is designed to aggregate the local details and global structural information of different modes of infrared-visible light,so as to avoid the limitation of a single mo-dality.The strategy is mainly composed of two global feature aggregation modules,a channel adaptive attention module and an auxiliary connection module,which realizes the integration of global structure and local detail information,fully learns the high-frequency detail in-formation in the image,and realizes effective feature alignment and information complementation,so as to improve the learning ability of complete features in the image fusion process.An autoencoder model is used combining convolutional neural networks and Transformer.Compared with the classic multi-source image fusion method on multi-source data sets,the proposed algorithm has achieved excellent results in image fusion.关键词
卷积神经网络/Transformer/多模态图像/多模态图像融合/自编码器Key words
convolutional neural networks/Transformer/multi-source image/multi-source image fusion/autoencoder分类
信息技术与安全科学引用本文复制引用
于城香,李会荣..基于复合注意融合策略的多源图像融合算法研究[J].计算机技术与发展,2025,35(11):69-77,9.基金项目
陕西省自然科学基础研究计划项目(23JSY048) （23JSY048）
商洛学院科研基金项目(24SKY023) （24SKY023）
