Enhancing Khalifa Satellite Imagery Resolution with AI-Powered Super Resolution Generative Adversarial Networks (SRGAN)

KhalifaSat is a multispectral remote sensing Earth observation satellite used for various analyses such as building detection and road extraction. However, the limited resolution of satellite images can result in blurry details. For this limitation to be removed, super resolution techniques have been introduced, with Generative Adversarial Networks (GANs) being a notable approach. This research proposes a novel algorithm called Hybrid SRGAN, which simplifies the Super-Resolution Generative Adversarial Network (SRGAN) model while maintaining high accuracy. The proposed modifications involve removing batch normalization layers and utilizing bicubic interpolation for resizing low-resolution images. By eliminating batch sampling, training speed and generalization across different image types are improved. To assess the performance of the suggested model by using the metrices like SSIM, PSNR, and BRISQUE, showcasing its superior reconstruction quality and overall performance, particularly in the luminance channel(YCbCr), when compared to cutting-edge algorithms.