A Benchmark Study on Image Quality Assessment and Interpretability in Super-Resolution

Abstract

Super-resolution (SR) is a vital technique in fields such as medical imag ing, surveillance, and satellite observation. However, evaluating the quality of super- resolved images remains challenging, especially without reference images. This thesis focuses on No-Reference Image Quality Assessment (NR IQA) for SR outputs, providing a detailed review of existing methods, their principles, strengths, limitations, and real-world applications. Additionally, the integration of Explainable Artificial Intelligence (XAI) offers insights into the decision-making processes of SR models, enhancing their transparency and reliability. This work contributes to the development of more effective and interpretable SR systems.