We propose a robust and general image fusion framework that requires no additional training, while effectively adapting to diverse fusion scenarios and effectively addressing various forms of interference.