SAR Image Super-Resolution based on Artificial Intelligence
Wei Yang, Ziqian Ma, Yingru Shi · IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium · 2022
High-resolution synthetic aperture radar (SAR) image can provide detailed information of the target, which is a benefit for improving the performance of the following interpretation application. However, the higher the resolution, the more complex the system and the higher the cost. A new challenge is how to obtain high-resolution target images from medium resolution images, by using new technology, such as artificial intelligence. In this paper, a new method is proposed to improve the resolution based on the SRGAN-SSIM. Since SRGAN is developed for nature image super-resolution, which results in a poor performance for SAR image, pre-processing is implemented. A modified Non-Local Means (NLM) is adopted for speckle noise suppression. Then, SRGAN based net is used for super-resolution, and the loss function is optimized according to the SSIM. Finally, the method is verified by Terra-SAR images.