Fig. 1. The framework of interactive self-attention fusion method

Fig. 2. The frameworks of Token ViT， Channel ViT and vision transformer

Fig. 3. The subjective comparison results of different fusion models

Fig. 4. The subjective comparison of “Nato_camp”， “Kaptrin_1123”and “Street” selected from the TNO dataset

Fig. 5. The objective comparison results of different methods for the TNO dataset

Fig. 6. The subjective comparison of “00443”and “03989” selected from the M³FD dataset

Fig. 7. The objective comparison results of different methods for the M³FD dataset

Fig. 8. The subjective comparison of “FILR_08910” and “FILR_06307” selected from the Roadscene dataset

Fig. 9. The objective comparison results of different methods for the Roadscene dataset