Infrared and visible image fusion based on a two-stage fusion strategy and feature interaction block

Infrared and visible image fusion technology (IVIFT) can combine the advantages of infrared and visible imaging systems and reduce the influence of particular environments, such as snow, darkness, fog, etc. Therefore, IVIFT is widely applied in security inspection, night monitoring, and remote sensing. However, many existing methods utilize a single-stage approach to optimize the model, which often causes weak robustness and an imbalance between the intensity and detail information. To solve this issue, we propose an infrared and visible image fusion method based on a two-stage fusion strategy and a feature interaction block (TFfusion). Specifically, a two-stage fusion strategy is developed to balance the salient target and the texture information retaining. The texture information is fused in Stage I, the salient targets are fused in Stage II, and Stage I guides Stage II in extracting texture information. A feature interaction block is designed to enhance the correlation between the source images and the fused image by sharing the features with each other. Quantitative and qualitative experiment results demonstrate that TFfusion achieves competitive performance and strong robustness in fusing the infrared and visible images compared with other advanced fusion methods.