ESRGAN - Based Super Resolution Model for Image Enhancement

Low-resolution X-ray images often hinder accurate diagnosis due to missing fine anatomical details. This project presents a novel application of Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN) to improve the clarity of such images. By leveraging perceptual, adversarial, and content loss functions, the proposed model reconstructs high-frequency details without introducing artifacts or blurring. The workflow integrates pre-processing with OpenCV, resolution enhancement via ESRGAN, and post-processing for quality refinement. Unlike traditional upscaling methods such as bicubic interpolation, this model preserves critical features like bone structures and soft tissues, ensuring better diagnostic utility. A user-friendly web interface further enables real-time enhancement, especially beneficial in under-resourced healthcare settings. The project is aligned with Sustainable Development Goals (SDGs), particularly Good Health and Well-being, Innovation and Infrastructure, and Reduced Inequalities, by democratizing access to AI-powered medical imaging tools. Initial tests indicate improved performance over baseline methods, making it a promising solution for medical diagnostics. Future work includes broader format support, batch processing, and extension to other modalities like CT and MRI.