The advancement of sensor technology has made it possible to capture scenes from several imaging modes. Combining the rich texture and color information from visible light photos with the comprehensive thermal radiation information from infrared photographs, a technique known as visible-infrared image fusion (VIF) improves both machine vision and human visual perception. Nevertheless, current fusion techniques frequently overlook frequency information and fall short of accurately evaluating the significance of visible and infrared images in various contexts. We propose a spatial-frequency interaction VIF framework to address these problems. This framework uses a weight allocation network to adaptively assign importance to visible and infrared images based on scene characteristics. It also records both local structural and global frequency information. Experiments conducted on four VIF datasets show that our approach outperforms recently developed sophisticated algorithms, especially improving edge detail and texture information in nighttime situations.