The proposed framework provides a robust, scalable solution for real-world illumination enhancement across diverse lighting conditions and consistently outperforms state-of-the-art supervised and unsupervised methods in terms of fidelity, perceptual quality, and generalization.
Abstract
Backlit and low-light images often suffer from severe exposure imbalance or global underexposure, presenting significant challenges for both visual perception and downstream computer vision tasks. In this paper, we propose a unified, unsupervised enhancement framework that addresses both types of degradation without relying on paired ground-truth data. Our approach builds on CLIP-guided prompt learning to semantically supervise enhancement using learned positive and negative textual prompts. To improve the quality of our improvements over prior work, we design a symmetric residual U-Net backbone augmented with an Atrous Spatial Pyramid Pooling module. This architecture captures multi-scale contextual information, enabling adaptive correction under spatially heterogeneous illumination. During training, the enhancement network is guided by CLIP-based semantic similarity losses and refined via an iterative prompt optimization mechanism. Extensive experiments on both paired and unpaired datasets, including BAID, Backlit300, LOL, and VE-LOL-L, demonstrate that our framework consistently outperforms state-of-the-art supervised and unsupervised methods in terms of fidelity, perceptual quality, and generalization. Furthermore, our work emphasizes the need for stronger benchmarking protocols for backlit enhancement, a relatively underexplored area. The proposed framework provides a robust, scalable solution for real-world illumination enhancement across diverse lighting conditions.
Low-light images often suffer from severe noise, low contrast. Traditional low-light enhancement methods rely on paired data or fixed priors, limiting generalization in real world scenarios. To address these problems, we propose a Progressive Prompt-Guided Enhancement Network (PPGENet) for unsupervised low-light image enhancement. Unlike previous CLIP-based methods that rely on fixed or single-stage prompts, directly applying CLIP to low-light enhancement faces challenges in achieving progressive quality improvements. By leveraging multimodal priors from CLIP, our method learns adaptive prompts that progressively guide the enhancement process. We introduce a three-stage framework. First, the prompt warm-up stage initializes learnable prompts by performing cross-entropy classification on mixed low-light and normal light images, thereby establishing semantic anchors in CLIP space. Second, the unsupervised reconstruction stage trains a Unet enhancement network guided by the fixed prompts, incorporating multiple loss functions to restore brightness and suppress noise. Third, the prompt refinement stage progressively fine tunes the prompts using pseudo-labels derived from intermediate enhancements and dynamic margin ranking loss, thereby enforcing progressive semantic ordering. This framework alternates between prompt refinement and network training until convergence. Extensive experiments on multiple datasets demonstrate that PPGENet achieves superior performance in both quantitative metrics and visual quality.
Mingtong Chen, Xiaowen Shi, Yongqiang Tang et al.· International Conference on...· 0 citations
Image enhancement is a widely researched area in the domain of computer vision, particularly image processing. Among the subdomains, low-light image enhancement (LLIE) receives considerable attention due to problems and challenges imposed by poor lighting conditions. As such, low-light images suffer from poor visibility, distorted colors, and loss of details, which limits their usability in many applications. The traditional methods have struggled to preserve such details and make the images susceptible to over-enhancement. Whereas, the learning-based techniques rely heavily on paired datasets for training. Therefore, we propose a Bidirectional Conv-GRU integrated GAN framework. Involving bidirectional Conv-GRU modules in our use-case enables the model to capture both local textures and long-range feature dependencies. Also, the use of unpaired datasets allows it to learn flexible and realistic mappings without the strict need for aligned image pairs. The results demonstrate the potency of our proposed work as compared to state-of-the-art methods.
Palak Deb Patra, Santosh Kumar Panda, Manoj Kumar Bishwal et al.· International Conference on...· 0 citations
Low-light image enhancement is crucial in situations where visible sensors might suffer from severe noise and information loss ( e.g., nighttime surveillance). Recent approaches investigate auxiliary modalities invariant to illumination to improve the performance, such as thermal infrared imaging. We propose a Multimodal Intrinsics-Guided Framework that integrates RGB and thermal data to reconstruct well-lit images. Our method utilizes a two-stage pipeline: first, we employ an intrinsic decomposition strategy to separate re-flectance and shading components through knowledge distillation, where a teacher network guides a student model in re-constructing consistent intrinsic components; then, a refine-ment stage restores fine structures and visual details. We train the proposed model on synthetic data from HDRT dataset and demonstrate strong generalization to real-world benchmarks such as LLVIP and V-TIEE, outperforming state-of-the-art methods in most evaluation metrics. Code is available at : https://github.com/simonemelc/TIRGlow
S. Melcarne, J. Dugelay· International Conference on...· 0 citations
Low-light image enhancement aims to improve visual visibility and perceptual quality under challenging illumination conditions. However, conventional convolutional neural networks (CNNs) are inherently limited in modeling long-range dependencies due to their restricted receptive fields, which often leads to insufficient global context modeling and suboptimal restoration results. To address this limitation, we propose MSHCDI-Net, a Multi-Scale Hybrid Cross-Domain Interaction Network that effectively integrates CNN and Transformer branches to jointly capture local texture details and global contextual relationships. Specifically, the proposed framework adopts a hierarchical encoder–decoder architecture to perform multi-scale feature extraction and progressive reconstruction. A cross-domain interaction mechanism is introduced to facilitate effective information exchange between convolutional and Transformer representations across multiple resolutions, enabling complementary modeling of fine-grained structures and long-range dependencies. Through adaptive feature fusion and multi-scale guidance, the network achieves improved structural consistency and detail restoration in low-light scenes. Extensive experiments on several public benchmarks demonstrate the effectiveness of the proposed method. MSHCDI-Net achieves 23.45 dB PSNR / 0.848 SSIM on LOL-v1, 23.74 dB / 0.910 SSIM on LOL-v2-synthetic, and 22.24 dB / 0.868 SSIM on LOL-v2-real, demonstrating competitive performance in both quantitative metrics and visual quality.
Bin Chen, Peitao Li, Chaobing Zheng et al.· PLoS ONE· 0 citations
Existing methods typically require training a separate model for each dataset, making them difficult to generalize across diverse illumination conditions. To address this limitation, we propose a novel low-light image enhancement method based on a Mixture of Experts (MoE) mechanism with fast adaptation. In our framework, the MoE gating network adaptively fuses the outputs of multiple experts to handle different lighting conditions, while only the expert and gating networks are fine-tuned when adapting to new datasets, significantly improving training efficiency and generalization. Each expert is designed as a multi-task module that jointly performs color correction and noise reduction, thereby enhancing both visual fidelity and robustness. Extensive quantitative and qualitative experiments demonstrate that the proposed method not only surpasses state-of-the-art approaches in noise reduction and color preservation, but also rapidly adapts to new illumination distributions with fast training across multiple benchmark datasets with significantly reduced fine-tuning cost and training time.
Yi Wang, Haonan Su, Zhaolin Xiao· IEEE Signal Processing Lette...· 0 citations
This work proposes a model-driven deep neural network to effectively handle the joint degradation of low light and blur and designs an illumination enhancement module (IEM) and a reflectance refinement module (RRM) to improve brightness, restore fine details, and suppress noise.
Yao Xiao, Youshen Xia, Zhenyu Lu et al.· IEEE Transactions on Neural...· 0 citations