<p>Infrared and visible image fusion aims to extract complementary multimodal information and integrate it into a single image for more comprehensive scene representation. Despite research progress, existing fusion frameworks remain constrained by differences in imaging mechanisms, limiting their ability to capture multidimensional feature information and achieve precise feature alignment. To address this issue, this paper proposes MAFIFusion, a multi-attention and feature interaction based on image fusion network. In this approach, guided by different lighting information, a multi-attention mechanism fusion module based on differential perception (MAM-DP) is designed. This module cascades the multi-head channel self-attention mechanism (MCSAM), index-based spatial attention (ISA), and corner-based global attention (CGA), enabling the model to efficiently extract both global and local features. Additionally, a feature interaction fusion module (FIFM) is introduced, comprising spatial domain feature interaction and channel domain feature interaction, to further perform structural alignment and semantic enhancement, ensuring the integrity of complementary information in the fused images. Extensive experimental results across five public datasets demonstrate that our algorithm outperforms existing methods in overall performance, exhibits strong adaptability to complex scenes, and generates fusion results that enrich scene detail and information content.</p>