Infrared and visible image fusion using LST and VSM-GF under PDE decomposition

The fusion of infrared and visible images integrates complementary information from both modalities, enhancing image quality in various applications. This paper presents a novel fusion method based on Partial Differential Equations (PDEs) decomposition, Local Statistical Texture (LST) model, and Visual Saliency Mapping-Guided Filtering (VSM-GF). PDEs are employed to decompose source images into base and detail layers. LST is utilized to generate adaptive weight maps for detail layers, while VSM-GF enhances the structural consistency in base layers. Extensive experiments were conducted on four publicly available datasets including TNO, LLVIP, M3FD, and RoadScene, and the proposed algorithm was quantitatively compared with nine traditional and deep learning-based fusion approaches on six evaluation metrics, including PSNR, MSE, $Q_{abf}$, SSIM, MS-SSIM, and FMI_pixel. Experimental results show that the proposed algorithm effectively preserves fine details and reduces artifacts, while also demonstrating superior performance across most quantitative metrics.