Infrared and visible images fusion network based on NestFusion

In this paper, we propose a novel infrared and visible light fusion network, referred to as the Dense Nested Fusion Network (DNFN). The nested connection-based architecture is capable of preserving a substantial amount of information from the input data from a multi-scale perspective. The proposed method comprises three key components: the encoder, fusion strategy, and decoder. We incorporate both spatial attention and channel attention models as the fusion strategy, which effectively captures the importance of each spatial location and each channel with deep features. Experiments were conducted on publicly available datasets.