Infrared and visible image fusion based on fourth order partial differential equation and cross bilateral filter

Fusion of infrared and visible images can integrate the complementary information in the source images to generate high-quality fused images that can highlight the prominent targets and contain rich texture details. However, the existing image fusion methods have the problem of information loss at the edge of the target. In order to solve this problem, we propose a new fusion method. Firstly, we utilize the cross bilateral filter (CBF) to extract both the detail layer and base layer information from the image. We tackle this by solving a fourth order partial differential equation (FPDE) to derive weights for the detail layer information. Then we employ principal component analysis (PCA) to obtain optimal weights, which are used to fuse the detail layer image. Then we use average fusion rule to merge the base layer information. Finally, the base layer and detail layer images are combined to generate the final image. We conduct experiments in multiple scenarios on public datasets and the results demonstrate that the fused images produced by our proposed method exhibit clear edges. Meanwhile fusion images achieved excellent performance in both quantitative and qualitative assessments.