Answers are grounded in this paper and cite their sources.
Infrared and visible image fusion (IVIF) is crucial for visual perception; however, balancing global thermal radiation information and local texture details remains challenging. To address this issue, this paper proposes FDSFusion, a dual-domain initialization and structure-driven refinement network for infrared–visible image fusion. Specifically, a spectral–spatial context block is designed to initialize shallow representations from both the spectral and spatial domains, thereby enhancing global contextual perception while preserving local details. In addition, a structure-guided texture refinement (SGTR) module is introduced, which uses structural features as semantic guidance to suppress cross-modality noise and refine texture details through inverse cross-attention. Experiments on the TNO and RoadScene demonstrate that FDSFusion effectively preserves structural consistency and texture details.
No takes yet. Share an insight, caveat, or question.
han et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: