CANDLE: Illumination-Invariant Semantic Priors for Color Ambient Lighting Normalization

arXiv cs.CV / 4/6/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses the difficulty of color ambient lighting normalization under multi-colored illumination, where chromatic shifts and material-dependent reflectance prevent reliable recovery of intrinsic object color using existing priors.
It proposes CANDLE, leveraging DINOv3 self-supervised features as illumination-robust semantic priors to counter illumination-induced chromatic bias.
CANDLE introduces DINO Omni-layer Guidance (D.O.G.) to inject multi-layer DINOv3 features into successive encoder stages, aiming to preserve semantic consistency across colored-light inputs.
It also uses a decoder-side color-frequency refinement (BFACG + SFFB) to suppress chromatic collapse and reduce detail contamination during reconstruction.
Experiments on CL3AN report a +1.22 dB PSNR improvement over the strongest prior, and challenge results show strong generalization (notably 3rd on NTIRE 2026 ALN Color Lighting and 2nd on fidelity in the White Lighting track) with code released on GitHub.

Abstract

Color ambient lighting normalization under multi-colored illumination is challenging due to severe chromatic shifts, highlight saturation, and material-dependent reflectance. Existing geometric and low-level priors are insufficient for recovering object-intrinsic color when illumination-induced chromatic bias dominates. We observe that DINOv3's self-supervised features remain highly consistent between colored-light inputs and ambient-lit ground truth, motivating their use as illumination-robust semantic priors. We propose CANDLE (Color Ambient Normalization with DINO Layer Enhancement), which introduces DINO Omni-layer Guidance (D.O.G.) to adaptively inject multi-layer DINOv3 features into successive encoder stages, and a color-frequency refinement design (BFACG + SFFB) to suppress decoder-side chromatic collapse and detail contamination. Experiments on CL3AN show a +1.22 dB PSNR gain over the strongest prior method. CANDLE achieves 3rd place on the NTIRE 2026 ALN Color Lighting Challenge and 2nd place in fidelity on the White Lighting track with the lowest FID, confirming strong generalization across both chromatic and luminance-dominant illumination conditions. Code is available at https://github.com/ron941/CANDLE.