VPCI: Self-Supervised Visual Prompt-Guided Cross-Domain Interactive Image Fusion Framework

Image fusion combines information from multi-modality images to produce high-quality fused images with enhanced clarity, contrast, and informativeness. However, limited ground truth fusion data lead to difficulties in effectively training these fusion models. Moreover, current studies lack of fine-grained, domain style-specific guidance during image fusion processing. To address these issues, we propose a self-supervised Visual Prompt-guided Cross-Domain Interactive image fusion framework (VPCI), which introduces a Visual Prompt Generation (VPG) strategy that embeds visual features from different modalities as prompts, guiding the model to learn complementary information. Furthermore, to better integrate modality-specific styles, we introduce a Cross-domain Prompt-guided Interaction (CPI) module to generate information-rich fusion images by fully interacting with visual prompts. Experimental results show that our framework outperforms existing state-of-the-art methods. The code is available at https://github.com/liwead/VPCI.