Infrared and visible image fusion aims to integrate thermal radiation and visual texture into a single image, enhancing perception in challenging scenarios. Existing methods often suffer from imbalanced feature extraction, limited cross-modal synergy, and loss of fine-grained details during propagation. To address these challenges, we propose SNeFusion, a graph neural network (GNN)-based framework with enhanced node and edge generation for improved information interaction. We design a dynamic node generation module that integrates Adaptive Matrix Learning (AML) and a “small-to-large receptive field” progressive aggregation strategy, enabling fine-grained detail preservation while maintaining global structural coherence. Also we design a novel edge generation module leveraging residual mixing and Efficient Spatial Scanning 2D (ES2D) combined with a Sigmoid-gated mechanism, which optimizes gradient consistency and cross-modal alignment. Extensive experiments on MSRS and FLIR datasets demonstrate that our SNeFusion outperforms state-of-the-art methods in both qualitative visual quality and quantitative metrics.