KDFuse: A high-level vision task-driven infrared and visible image fusion method based on cross-domain knowledge distillation

Chenjia Yang, Xiaoqing Luo, Zhancheng Zhang, Zhiguo Chen, Xiao-jun Wu

Abstract: Highlights•Introduce knowledge distillation to address the collaboration between high-level vision task and image fusion task.•The multi-domain interaction distillation module and the multi-scale semantic perception module are designed to efficient extract and fuse semantic and fusion features respectively.•The Fourier transform is introduced into the loss function to help the model retain texture information better.