Self-Supervised Image Super-Resolution Quality Assessment based on Content-Free Multi-Model Oriented Representation Learning

[Paper (preprint)- arXiv] [Dataset - Zenodo]

Overview

This repository provides source code, dataset, and material associated for our 2026 paper titled "Self-Supervised Image Super-Resolution Quality Assessment based on Content-Free Multi-Model Oriented Representation Learning." Our paper proposes the first self-supervised learning approach to assess the quality of super-resolution images generated from real low-resolution inputs. S3RIQA outperforms no-reference SR-IQA metrics in existing benchmarks using multiple SR-IQA datasets, with realistic LR images.

Authors

• Kian Majlessi
• Amir Masoud Soltani
• Mohammad Ebrahim Mahdavi
• Aurelien Gourrier
• Peyman Adibi

Citation

If you find this project useful, then please consider citing both our paper and dataset.

@article{majlessi2026s3riqa,
  title={Self-Supervised Image Super-Resolution Quality Assessment based on Content-Free Multi-Model Oriented Representation Learning},
  author={Majlessi, Kian and Soltani, Amir Masoud and Mahdavi, Mohammad Ebrahim and Gourrier, Aurelien and Adibi, Peyman},
  journal={arXiv preprint arXiv:2602.10744},
  year={2026}
}

@dataset{majlessi2026srmorss
  title={SRMORSS: Super-Resolution Model-Oriented Realistic Self-Supervision Dataset},
  author={Majlessi, Kian and Soltani, Amir Masoud and Mahdavi, Mohammad Ebrahim and Gourrier, Aurelien and Adibi, Peyman},
  publisher={Zenodo},
  version={1.0.0},
  url={https://doi.org/10.5281/zenodo.18479156},
  doi={10.5281/zenodo.18479156},
  year={2026},
}

Code Setup

Coming Soon!

Source Code License

Dataset License

Acknowledgments

This work has benefited from a French government grant managed by the Agence Nationale de la Recherche under the France 2030 program, reference ANR-23-IACL-0006.