InterLight: Leveraging Intrinsic Illumination Priors for Low-Light Image Enhancement

2026-05-19

AI Summary

This work addresses the challenges of low-light image enhancement—namely, low contrast, detail loss, prominent noise, and the tendency of existing methods to produce over-enhancement and color distortion—by proposing a novel illumination-aware enhancement framework. The framework systematically exploits and operationalizes sensor-level intrinsic illumination priors for the first time, integrating physics-guided data augmentation, scene-adaptive illumination cue modeling, and a luminance-gated intrinsic memory mechanism to selectively compensate for information loss. Furthermore, it incorporates self-supervised consistency regularization to facilitate learning of illumination-invariant features. Built upon Retinex theory, the proposed deep architecture significantly improves texture recovery quality and visual consistency across multiple benchmarks while effectively suppressing over-enhancement and color distortion.

Abstract

Low-Light Image Enhancement (LLIE) has long been a challenging problem in low-level vision, as insufficient illumination often leads to low contrast, detail loss, and noise. Recent studies show that deep learning-based Retinex theory can effectively decouple illumination and reflectance. However, existing methods frequently suffer from over-enhancement or color distortion, and often assume uniform noise or ideal lighting. To address these limitations, we propose InterLight, a novel framework that systematically excavates and operationalizes intrinsic illumination priors for LLIE.Our core insight is that robust enhancement requires not just estimating illumination, but constructing an illumination-aware pipeline. We first inject sensor-level illumination-response priors via physics-guided augmentation, then represent the degradation through adaptive prompts conditioned on the scene's latent illumination state. This explicit representation directly guides a luminance-gated intrinsic memory mechanism to selectively compensate for information loss, prioritizing reconstruction in dark regions while preserving fidelity in bright ones. Finally, the entire process is regularized by a self-supervised consistency objective that distills illumination-invariant features. By deeply exploiting intrinsic illumination priors, our method achieves clearer textures and more visually coherent enhancement results. Extensive experiments across multiple benchmarks demonstrate the effectiveness of our approach. Code is available at: https://github.com/House-yuyu/InterLight.

Problem

• Low-Light Image Enhancement
• Illumination Priors
• Over-enhancement
• Color Distortion
• Image Degradation

Innovation

• intrinsic illumination priors
• physics-guided augmentation
• adaptive prompts
• luminance-gated memory
• self-supervised consistency