Visible and infrared image fusion integrates data from different sensors to produce images with richer information. The core goal is to preserve the unique thermal features of infrared images while maintaining the fine textures present in visible light images. However, fusion often encounters two key challenges: ringing artifacts together with texture loss. Ringing distorts edges, whereas texture loss diminishes image detail. To address these challenges, we propose a novel image fusion framework that consists of three main modules: VI-Fuser, Pseudo-Sensing, and HSNet. The VI-Fuser module is designed to extract and integrate complementary features from visible and infrared images, thereby enhancing the representational capacity of the fused features. The Pseudo-Sensing module enables self-supervised learning by generating pseudo-labels in the absence of real fused ground truth, thus facilitating robust optimization of the fusion network. The HSNet module incorporates a High-order Degradation model, named HD, a U-Net discriminator with Spectral Normalization, called U-SNet and dense blocks. Within this module, the HD component effectively reduces ringing and overshooting artifacts to enhance edge clarity, while U-SNet improves the preservation of texture details, leading to more refined and reliable feature representations. Experimental results demonstrate that our model effectively preserves thermal targets along with fine structural information, yielding images with better overall integrity along with richer details.