Super-Resolution Reconstruction of UAV Images with GANs: Achievements and Challenges

Images taken by UAVs are crucial in many applications due to their ability to provide accurate and thorough information. They are used in various sectors such as agriculture, environmental monitoring, and geology. Although UAVs have advantages, their images may still suffer from issues like blurriness and lower resolution caused by hardware constraints, altitude, and distance. Improving the quality and sharpness of images by using image super-resolution reconstruction is a challenging obstacle in the realm of computer vision, as it requires the transformation of low-resolution images into high-resolution ones. After generative adversarial networks (GANs) were introduced, some models showed promising results in improving image clarity. In this study, we introduce a super resolution GAN (SRGAN) model to improve the resolution of UAV images. By using metrics like Peak Signal-to-Noise Ratio (PSNR) and structural similarity index (SSIM), we found that good results in GAN models can be achieved even with a small dataset. The results show an SSIM score of 0.52 and a PSNR score of 30.2 after 8000 epochs. It is recommended to use advanced techniques to enhance SRGAN model performance.