Abstract:Low-light images incur several complicated degradation factors such as poor brightness, low contrast, color degradation, and noise. Most previous deep learning-based approaches, however, only learn the mapping relationship of single channel between the input low-light images and the expected normal-light images, which is insufficient enough to deal with low-light images captured under uncertain imaging environment. Moreover, too deeper network architecture is not conducive to recover low-light images due to extremely low values in pixels. To surmount aforementioned issues, in this paper we propose a novel multi-branch and progressive network (MBPNet) for low-light image enhancement. To be more specific, the proposed MBPNet is comprised of four different branches which build the mapping relationship at different scales. The followed fusion is performed on the outputs obtained from four different branches for the final enhanced image. Furthermore, to better handle the difficulty of delivering structural information of low-light images with low values in pixels, a progressive enhancement strategy is applied in the proposed method, where four convolutional long short-term memory networks (LSTM) are embedded in four branches and an recurrent network architecture is developed to iteratively perform the enhancement process. In addition, a joint loss function consisting of the pixel loss, the multi-scale perceptual loss, the adversarial loss, the gradient loss, and the color loss is framed to optimize the model parameters. To evaluate the effectiveness of proposed MBPNet, three popularly used benchmark databases are used for both quantitative and qualitative assessments. The experimental results confirm that the proposed MBPNet obviously outperforms other state-of-the-art approaches in terms of quantitative and qualitative results. The code will be available at https://github.com/kbzhang0505/MBPNet.

Learning to Adapt to Light

Mutual Support and Promotion: Learning Structure Compensation and Context Completion for Low-Light Vision

Learning Adaptive Lighting via Channel-Aware Guidance

Progressive feature-aware recurrent net for low-light image enhancement

LightingNet: An Integrated Learning Method for Low-Light Image Enhancement

Unsupervised Illumination Adaptation for Low-Light Vision

Dark2Light: multi-stage progressive learning model for low-light image enhancement

Brighten-and-Colorize: A Decoupled Network for Customized Low-Light Image Enhancement

Self-Aligned Concave Curve: Illumination Enhancement for Unsupervised Adaptation

Learning an adaptive model for extreme low‐light raw image processing

Learning an Adaptive Model for Extreme Low-light Raw Image Processing

Toward Robust and Efficient Low-Light Image Enhancement:Progressive Attentive Retinex Architecture Search

Beyond Brightening Low-light Images

Generalized Lightness Adaptation with Channel Selective Normalization

Multi-Branch and Progressive Network for Low-Light Image Enhancement

Joint Correcting and Refinement for Balanced Low-Light Image Enhancement

LACN: A lightweight attention-guided ConvNeXt network for low-light image enhancement

Division gets better: Learning brightness-aware and detail-sensitive representations for low-light image enhancement

Lit me up: A reference free adaptive low light image enhancement for in-the-wild conditions

Low-Light Image Enhancement by Learning Contrastive Representations in Spatial and Frequency Domains

Attention Guided Low-Light Image Enhancement with a Large Scale Low-Light Simulation Dataset