MM-CGAN: a multi-module conditioned generative adversarial network for sar image to optical image translation
Feifei Dong, Chisheng Wang · Remote Sensing Letters · 2025
Synthetic Aperture Radar (SAR) is capable of all-weather operation, unaffected by atmospheric conditions, and offers a much higher image resolution compared to traditional optical images, which are easily disrupted by environmental factors. However, optical images typically provide a wealth of visual information, making it significant to research how SAR images can be converted into optical images. In recent years, with the continuous development of deep learning, using Generative Adversarial Networks (GANs) to transform SAR images into optical images has become mainstream. This paper proposes a multi-module conditional generative adversarial network (MM-CGAN), which incorporates three different modules to enhance the detail in the generated optical images. It also employs a multi-scale discriminator to strengthen the network’s discriminative ability. Finally, comparisons with several classic translation models on public datasets have shown that our model achieves higher scores in image quality assessment, demonstrating its superiority.