Infrared and visible image fusion based on cross-modality feature alignment

The goal of infrared and visible image fusion is to integrate complementary information from infrared and visible images of the same scene to generate a high-quality composite image that preserves the advantages of both modalities. Although many existing infrared and visible image fusion methods have achieved promising results, they still suffer from the coupling of high- and low-frequency information. To address this issue, this paper proposes a feature alignment-based fusion method that leverages feature decoupling and loss-guided alignment to ensure that the infrared and visible features not only capture shared semantic information but also retain their respective texture details. The proposed approach hierarchically separates image features and aligns corresponding layers under the guidance of carefully designed loss functions to achieve effective fusion. Experimental results demonstrate that, compared to state-of-the-art methods, the proposed method significantly improves both fusion quality and detail preservation across multiple datasets, providing higher-quality fused images for downstream image processing tasks.