MSR-net: Low-light Image Enhancement Using Deep Convolutional Network

By Liang Shen, Zihan Yue, Fan Feng, Quan Chen, Shihao Liu, Jie Ma. arXiv.org, 2017

313 citations recorded by the source corpus.

Images captured in low-light conditions usually suffer from very low contrast, which increases the difficulty of subsequent computer vision tasks in a great extent. In this paper, a low-light image enhancement model based on convolutional neural network and Retinex theory is proposed. Firstly, we show that multi-scale Retinex is equivalent to a feedforward convolutional neural network with different Gaussian convolution kernels. Motivated by this fact, we consider a Convolutional Neural Network(MSR-net) that directly learns an end-to-end mapping between dark and bright images. Different fundamentally from existing approaches, low-light image enhancement in this paper is regarded as a machine learning problem. In this model, most of the parameters are optimized by back-propagation, while the parameters of traditional models depend on the artificial setting. Experiments on a number of challenging images reveal the advantages of our method in comparison with other state-of-the-art methods from the qualitative and quantitative perspective.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/6d32ca437e9d49a974ecaea74a5ba0f706689d81

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Benchmarking Low-Light Image Enhancement and Beyond (2021, background): “With the rise of deep learning-based methods (Lore et al. 2017; Shen et al. 2017; Wei et al. 2018), some new datasets provide paired low/normal-light images for training and eval-123 uating low-light enhancement methods from the perspective of low-level signal ﬁdelity.”
An Experiment-Based Review of Low-Light Image Enhancement Methods (2020, methodology): “Based on Retinex theory, Shen et al. [268] analyzed the performance of the MSR algorithm from the perspective of a CNN framework and proposed a method of enhancing low-light images by using an MSR network (MSR-net) based on a CNN architecture, while Guo et al. [269] proposed our pipeline neural…”
Learning Enriched Features for Fast Image Restoration and Enhancement (2022, background): “Notable works employ Retinex-inspired networks [4], [84], [85], [86],”
GLADNet: Low-Light Enhancement Network with Global Awareness (2018, methodology): “Different from [8][11] which synthesize pairs on 8-bit RGB images, we synthesize pairs on raw images.”
LR3M: Robust Low-Light Enhancement via Low-Rank Regularized Retinex Model (2020, background): “In [25], deeply root in multi-scale Retinex representation, a feed-forward convolutional neural network with different Gaussian convolution kernels is proposed to learn an end-to-end mapping between dark and bright images.”
Ultra-High-Definition Low-Light Image Enhancement: A Benchmark and Transformer-Based Method (2022, methodology): “(Shen et al. 2017) created an LLIE dataset of 10 , 000 image pairs, where 8 , 000 for training and 2 , 000 for testing.”
Learning Enriched Features for Real Image Restoration and Enhancement (2020, methodology): “roposed in the literature [8,74,54,83]. Recently, CNNs have been successfully applied to general, as well as low-light, image enhancement problems [52]. Notable works employ Retinex-inspired networks [89,107,124], encoder-decoder networks [18,68,81], and GANs [19,51,25]. 3 Proposed Method In this se”
R2RNet: Low-light Image Enhancement via Real-low to Real-normal Network (2021, methodology): “MSR-Net [23] utilized different Gaussian convolution kernels to learn the mapping of low/normal-light images.”
Low-Light Image Enhancement with Normalizing Flow (2021, methodology): “The (Shen et al. 2017) illustrates the close relationship between Retinex and CNN with Gaussian convolution kernels, two separated deep networks are used for decomposition in (Wei et al. 2018), and (Wang et al. 2019b) propose a progressive Retinex framework that the illumination and reﬂection maps…”
Sparse Gradient Regularized Deep Retinex Network for Robust Low-Light Image Enhancement (2021, methodology): “• Some recent data-driven methods learn a direct mapping between low-light and normal light images [14], [19].”
Attention Guided Low-Light Image Enhancement with a Large Scale Low-Light Simulation Dataset (2019, methodology): “MSRNet [68] learns an end-to-end mapping between dark/bright images by using different Gaussian convolution kernels.”
Getting to Know Low-light Images with The Exclusively Dark Dataset (2018, methodology): “In order to solve this noise issue, existing off-the-shelf denoising algorithms were commonly employed as a post-processing step ([17], [18], [49]) or by incorporating mechanisms in the proposed models ([13], [11]).”
Kindling the Darkness: A Practical Low-light Image Enhancer (2019, methodology): “Motivated by this, they constructed a convolutional neural network (MSR-net) [20] to learn an end-to-end mapping between dark and bright images.”
Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study (2019, background): “Learning based low-light image enhancement methods [22]–[24] have also been studied, where low-light images used for training are synthesized by applying random Gamma transformation on natural normal light images.”
Beyond Brightening Low-light Images (2021, methodology): “Shen et al. (2017) deemed that multi-scale Retinex is equivalent to a feed-forward convolutional neural network with different Gaussian convolution kernels.”
From Fidelity to Perceptual Quality: A Semi-Supervised Approach for Low-Light Image Enhancement (2020, methodology): “Later on, various methods based on different network designs [26, 3, 24, 27, 30] are proposed.”