Adversarial network for unsupervised infrared image colorization based on full-scale feature fusion and cosine contrastive learning

Tingting Liu, Yujue Cai, Guiping Chen, Hongguang Wei, Junqi Bai, Yuan Liu, Xiubao Sui, Qian Chen

Abstract: Highlights•A novel method (CCLGAN) is proposed for unsupervised infrared image colorization task.•A novel cosine contrastive loss is proposed to enhance color image generation quality.•A UNet-based generator with full-scale feature fusion and Mamba module is proposed.•The Mamba module incorporates 3D neural attention to enhance key features and suppress redundancy.