This paper aims to improve the resolution of space imagery with a better version of Super-Resolution Generative Adversarial Network (SRGAN). Normally, it is impossible to obtain images from satellites with reasonable resolutions. We proposed the application of Dense Residual Blocks and Progressive Growing of the Generator together in the architecture of SRGAN to achieve more refined training while progressively increasing the resolution of the images. Dense residual blocks enable more feature reuse and enhance information flow within the network, hence facilitating better learning and the capacity to learn many complex textures and spatial information. Progressive Growing of the generator enables the model to handle larger upscaling tasks much more smoothly by gradually increasing the generation complexity with the passage of the training process. Together, these advances significantly improve the stability in training, processing as well as the learning efficiency with enhanced images. The proposed enhanced SRGAN outperforms existing methods for generating high-resolution satellite images and is thus extremely applicable to real-world tasks requiring the reconstruction of fine details from low-resolution inputs.