Adaptive Fusion of Infrared and Visible Images for Ancient Bamboo and Wooden Slips Based on SSA

Ancient bamboo and wooden slips, as the primary written medium before the invention of paper in China, often suffer from severe text degradation due to prolonged burial in soil environments, making the characters difficult to recognize. To enhance the text and detailed information on these slips, this paper proposes an adaptive image fusion method based on the salient spatial attention mechanism, aiming to combine the advantages of infrared and visible imaging to generate high-quality images that preserve clear ink writing while maintaining material texture and color information. First, a specialized Salient Spatial Attention (SSA) mechanism has been designed, which enhances ink information in infrared images through adaptive feature selection. Second, based on the unique visual characteristics of the slips' material and ink writing, a multi-scale information measurement strategy is proposed to achieve intelligent fusion of features at different scales, ensuring balanced expression of material texture and ink details. Finally, an unsupervised loss function adapted to the texture features of the slips has been introduced, eliminating the dependency on missing ground-truth data while ensuring that the fusion results maintain the original visual features. Experiments have shown that the proposed method outperforms existing image fusion methods in both visual effects and quantitative metrics, particularly in preserving slip text details, material texture, and color fidelity.