Infrared and visible image fusion based on spatial correlation attention

Abstract

Infrared and visible image fusion plays a crucial role in enhancing scene understanding and improving the detection performance of infrared imaging systems. One major challenge in image fusion is the loss of fine-grained details from the source images during the fusion process. To tackle this issue, we introduce IVFSCA—an image fusion method based on spatially correlated attention. Specifically, we design a spatial correlation attention mechanism to preserve spatial dependencies during feature extraction and fusion. This helps reduce the loss of fine-grained information. We further develop a multi-source feature fusion network with a cross-residual structure that effectively integrates multi-scale features from infrared and visible images, producing high-quality and robust fusion representations. A structurally symmetric encoder–decoder architecture is also adopted to improve the robustness of feature extraction and image reconstruction. Extensive experiments on public benchmark datasets demonstrate that the proposed IVFSCA achieves competitive performance compared to state-of-the-art fusion methods in both visual quality and quantitative metrics.