Image Generation and Super Resolution using Custom GANs

In recent years, Generative Adversarial Networks (GANs) have emerged as a powerful framework for image synthesis and enhancement. This paper presents a hybrid approach that combines a Deep Convolutional GAN (DCGAN) for generating low-resolution lunar surface images and an Enhanced Super-Resolution GAN (ESRGAN) for refining these outputs into high-resolution, photorealistic visuals. Our proposed architecture leverages the robust feature extraction capability of DCGANs alongside the perceptual fidelity of ESRGANs to produce visually convincing lunar imagery. The results demonstrate significant improvements in both structural accuracy and visual sharpness, enabling a deeper exploration into data-driven lunar visualization. Quantitative metrics such as PSNR and SSIM validate the quality enhancement, and qualitative results affirm the model's efficacy in preserving fine details. This dual-GAN pipeline underscores the potential of integrating generative and superresolution models for scientific and artistic applications in space imagery.