RetinexDualV2 accepted to CVPR 2026 Workshops

RetinexDualV2 (Kishawy & Chen, 2026) grounds RetinexDual in the physics of each degradation. A Task-Specific Physical Grounding Module (TS-PGM) extracts degradation-aware priors, such as rain masks and dark channels, and these guide the Retinex decomposition through a Physical-Conditioned Multi-head Self-Attention (PC-MSA) mechanism. As a result, one architecture handles different degradations without task-specific structural changes. The same model placed 4th and 5th in two NTIRE 2026 challenges.

References

CVPRW
RetinexDualV2: Physically-Grounded Dual Retinex for Generalized UHD Image Restoration

Mohab Kishawy and Jun Chen

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Jun 2026

NTIRE 2026: 4th and 5th Place Abs Paper Code Bib

RetinexDualV2 ranked 4th in the NTIRE 2026 Day and Night Raindrop Removal Challenge and 5th in the NTIRE 2026 Joint Denoising and Low-Light Image Enhancement (JDLLIE) track.

We propose RetinexDualV2, a unified, physically grounded dual-branch framework for diverse Ultra-High-Definition (UHD) image restoration. Unlike generic models, our method employs a Task-Specific Physical Grounding Module (TS-PGM) to extract degradation-aware priors (e.g., rain masks and dark channels). These explicitly guide a Retinex decomposition network via a novel Physical-Conditioned Multi-head Self-Attention (PC-MSA) mechanism, enabling robust reflection and illumination correction. This physical conditioning allows a single architecture to handle various complex degradations seamlessly, without task-specific structural modifications. RetinexDualV2 demonstrates exceptional generalizability, securing 4th place in the NTIRE 2026 Day and Night Raindrop Removal Challenge and 5th place in the Joint Noise Low-light Enhancement (JNLLIE) Challenge. Extensive experiments confirm the state-of-the-art performance and efficiency of our physically motivated approach. Code is available at https://github.com/ErrorLogic1211/RetinexDual/tree/master/RetinexDualV2
@inproceedings{kishawy2026retinexdualv2, title = {{RetinexDualV2}: Physically-Grounded Dual Retinex for Generalized {UHD} Image Restoration}, author = {Kishawy, Mohab and Chen, Jun}, booktitle = {Proceedings of the {IEEE/CVF} Conference on Computer Vision and Pattern Recognition ({CVPR}) Workshops}, month = jun, year = {2026}, pages = {1989--1998}, }