Fig. 1. Multi-stage image fusion process

Fig. 2. Central difference convolution and angular difference convolution

Fig. 3. Horizontal difference convolution and vertical difference convolution

Fig. 4. The overall structure of mixed differential convolution

Fig. 5. Overview of EfficientViT^[14]. (a) Architecture of EfficientViT; (b) Sandwich layout block; (c) Cascaded group attention

Fig. 6. Examples of differential images. (a) Visible image. (b) Infared image. (c) Difference image

Fig. 7. Overall framework of network structure

Fig. 8. The framework of CNN branch

Fig. 9. The framework of Transformer-branch

Fig. 10. The framework of improved coordinate attention

Fig. 11. The framework of MCA

Fig. 12. The framework of CMCA attention

Fig. 13. Feature fusion layer based on multi-dimensional coordinate co-attention mechanism

Fig. 14. The framework of decoder

Fig. 15. Network framework of the training stage

Fig. 16. Fusion results of three typical pairs of images

Fig. 17. Fusion results of ablation experiment