Validate MDIRNET on real-world mixed-degradation data

Validate the Multi-Degradation Image Restoration Network (MDIRNET) on real-world datasets containing mixed image degradations, including datasets such as DND.

Background

MDIRNET is evaluated primarily on benchmark settings involving synthetic or separately specified noise, rain, and blur combinations. The paper explicitly states that validation on real-world mixed-degradation datasets has not yet been performed, leaving the model's performance under naturally occurring combinations of degradations unresolved.

References

The low-rank prior is less effective for severe large-kernel blur, and validation on real-world mixed-degradation datasets (e.g., DND) remains future work.

— MDIRNET: Multi-Degradation Image Restoration Network via Deep Unfolding  (2610.01655 - Nadeem et al., 1 Oct 2026) in Section Conclusion