LiT-fusion: lightweight transformer for infrared-visible image fusion

作者：Tingyu Zhu, Gang Wang, Yanxiao Kang, Jinyong Chen · 年份：2025 · DOI：10.1117/12.3087040 · 被引用次数：1 · 研究领域：Advanced Image Fusion Techniques、Image Enhancement Techniques、Advanced Neural Network Applications

Infrared and visible image fusion has remained a focal research area in computer vision due to inherent challenges including computational complexity and feature selection difficulties. This paper presents a lightweight fusion network based on simplified linear attention mechanisms that synergistically combines thermal information from infrared sensors with textural details from visible cameras, thereby creating comprehensive scene representations that operate effectively across diverse illumination conditions. Our model incorporates simplified linear attention blocks that significantly enhance the fusion network's performance. The proposed architecture encompasses feature enhancement modules, cross-modal feature interaction modules, and feature fusion layers to effectively preserve discriminative information from heterogeneous modalities. Experimental results demonstrate that our proposed method exhibits superior performance with strong competitive advantages.