Filters








52 Hits in 1.8 sec

EnlightenGAN: Deep Light Enhancement without Paired Supervision [article]

Yifan Jiang, Xinyu Gong, Ding Liu, Yu Cheng, Chen Fang, Xiaohui Shen, Jianchao Yang, Pan Zhou, Zhangyang Wang
2021 arXiv   pre-print
We propose a highly effective unsupervised generative adversarial network, dubbed EnlightenGAN, that can be trained without low/normal-light image pairs, yet proves to generalize very well on various real-world  ...  Deep learning-based methods have achieved remarkable success in image restoration and enhancement, but are they still competitive when there is a lack of paired training data?  ...  Thanks to the unpaired training, EnlightenGAN could be easily adapted into EnlightenGAN-N without requiring any supervised/paired data in the new domain, which greatly facilitates its real-world generalization  ... 
arXiv:1906.06972v2 fatcat:sq3oxeq6pzdjdjmlamszanym54

CERL: A Unified Optimization Framework for Light Enhancement with Realistic Noise [article]

Zeyuan Chen, Yifan Jiang, Dong Liu, Zhangyang Wang
2022 arXiv   pre-print
For the real low-light noise removal part, we customize a self-supervised denoising model that can easily be adapted without referring to clean ground-truth images.  ...  We present Coordinated Enhancement for Real-world Low-light Noisy Images (CERL), that seamlessly integrates light enhancement and noise suppression parts into a unified and physics-grounded optimization  ...  Recently, several enhancement models without paired supervision are proposed.  ... 
arXiv:2108.00478v2 fatcat:4kvlnqc4rfhytnwlwrzmaxmdiy

A Two-stage Unsupervised Approach for Low light Image Enhancement [article]

Junjie Hu, Xiyue Guo, Junfeng Chen, Guanqi Liang, Fuqin Deng, Tin lun Lam
2020 arXiv   pre-print
Recently, deep learning based methods have been proposed to enhance low light images by penalizing the pixel-wise loss of low light and normal light images.  ...  However, most of them suffer from the following problems: 1) the need of pairs of low light and normal light images for training, 2) the poor performance for dark images, 3) the amplification of noise.  ...  Unsupervised Based Methods Unsupervised based methods attempt to enhance low light images without pairs of low light and normal light images.  ... 
arXiv:2010.09316v2 fatcat:iyxm5dwxf5entpic7hkjsxn7sa

ReLLIE: Deep Reinforcement Learning for Customized Low-Light Image Enhancement [article]

Rongkai Zhang, Lanqing Guo, Siyu Huang, Bihan Wen
2021 arXiv   pre-print
To tackle these two challenges, this paper presents a novel deep reinforcement learning based method, dubbed ReLLIE, for customized low-light enhancement.  ...  Low-light image enhancement (LLIE) is a pervasive yet challenging problem, since: 1) low-light measurements may vary due to different imaging conditions in practice; 2) images can be enlightened subjectively  ...  More recent methods [8] focus on unsupervised LLIE which directly enlightens low-light images without any paired training data.  ... 
arXiv:2107.05830v1 fatcat:irqt24s4ercdjoe3lvxkvztx2y

Low-Light Image and Video Enhancement Using Deep Learning: A Survey [article]

Chongyi Li and Chunle Guo and Linghao Han and Jun Jiang and Ming-Ming Cheng and Jinwei Gu and Chen Change Loy
2021 arXiv   pre-print
Low-light image enhancement (LLIE) aims at improving the perception or interpretability of an image captured in an environment with poor illumination.  ...  Recent advances in this area are dominated by deep learning-based solutions, where many learning strategies, network structures, loss functions, training data, etc. have been employed.  ...  [2] for low-light video enhancement. Reinforcement Learning. Without paired training data, Yu et al. [25] learn to expose photos with reinforcement adversarial learning, named DeepExposure.  ... 
arXiv:2104.10729v3 fatcat:d72ioaximbdb3nbjnfqvjcmqte

Enhancing Low-Light Images in Real World via Cross-Image Disentanglement [article]

Lanqing Guo, Renjie Wan, Wenhan Yang, Alex Kot, Bihan Wen
2022 arXiv   pre-print
Though weakly-supervised or unsupervised methods can alleviate such challenges without using paired training images, some real-world artifacts inevitably get falsely amplified because of the lack of corresponded  ...  Furthermore, we collect a new low-light image enhancement dataset consisting of misaligned training images with real-world corruptions.  ...  E-mail: lanqing001@e.ntu.edu.sg, {rjwan, wenhan.yang, eackot, bihan.wen}@ntu.edu.sg. , and the enhanced results using (b) KinD [1] (fully supervised method), (c) Zero-DCE [2] without paired supervision  ... 
arXiv:2201.03145v1 fatcat:krqddjs4izfzhf6n4gssofdn4y

Towards Robust Low Light Image Enhancement [article]

Sara Aghajanzadeh, David Forsyth
2022 arXiv   pre-print
We use a supervised learning method, relying on a straightforward simulation of an imaging pipeline to generate usable dataset for training and testing.  ...  [21] propose a deep recursive band network (DRBN) which decomposes an image using paired data then reconstructs without paired data using perceptual quality driven adversarial learning.  ...  Related Work Deep learning methods now dominate low light image enhancement. A recent survey [8] gives a comprehensive overview of existing works.  ... 
arXiv:2205.08615v1 fatcat:2zpabvywkreqreamnz64uzcigq

Unsupervised Low-Light Image Enhancement via Histogram Equalization Prior [article]

Feng Zhang, Yuanjie Shao, Yishi Sun, Kai Zhu, Changxin Gao, Nong Sang
2021 arXiv   pre-print
Deep learning-based methods for low-light image enhancement typically require enormous paired training data, which are impractical to capture in real-world scenarios.  ...  Extensive experiments demonstrate that our method performs favorably against the state-of-the-art unsupervised low-light enhancement algorithms and even matches the state-of-the-art supervised algorithms  ...  Deep learning based Methods Deep learning-based methods have dominated the research of low-light image enhancement. Lore et al.  ... 
arXiv:2112.01766v1 fatcat:k3oslkc5k5birhlkpu4b6eamh4

Semantic-Guided Zero-Shot Learning for Low-Light Image/Video Enhancement [article]

Shen Zheng, Gaurav Gupta
2021 arXiv   pre-print
This paper proposes a semantic-guided zero-shot low-light enhancement network (SGZ) which is trained in the absence of paired images, unpaired datasets, and segmentation annotation.  ...  Secondly, we propose a recurrent image enhancement network to progressively enhance the low-light image with affordable model size.  ...  Unsupervised methods avoid the tedious work for preparing paired training images. EnlightenGAN [20] is the first low-light image enhancement method trained without paired data.  ... 
arXiv:2110.00970v4 fatcat:hwihf25gfjboplqyczewqihhcy

Learning Deep Context-Sensitive Decomposition for Low-Light Image Enhancement [article]

Long Ma, Risheng Liu, Jiaao Zhang, Xin Fan, Zhongxuan Luo
2021 arXiv   pre-print
Enhancing the quality of low-light images plays a very important role in many image processing and multimedia applications.  ...  According to different training patterns, we construct CSDNet (paired supervision) and CSDGAN (unpaired supervision) to fully evaluate our designed architecture.  ...  Performance Evaluation of CSDGAN Here, we executed the unpaired supervision manner by utilizing the training pairs presented in EnlightenGAN [14] , which includes 914 low-light and 1061 normal light images  ... 
arXiv:2112.05147v1 fatcat:yoa5pfeynjaubl5r2hc6bhgg7q

DALE : Dark Region-Aware Low-light Image Enhancement [article]

Dokyeong Kwon, Guisik Kim, Junseok Kwon
2020 arXiv   pre-print
In this paper, we present a novel low-light image enhancement method called dark region-aware low-light image enhancement (DALE), where dark regions are accurately recognized by the proposed visual attention  ...  Our method can estimate the visual attention in an efficient manner using super-pixels without any complicated process.  ...  Enlight-enGAN [11] performed low-light image enhancement in an unsupervised manner without paired data.  ... 
arXiv:2008.12493v1 fatcat:dzmzjr5jkbfr7cejegzff63nme

Unsupervised Low-light Image Enhancement with Decoupled Networks [article]

Wei Xiong, Ding Liu, Xiaohui Shen, Chen Fang, Jiebo Luo
2022 arXiv   pre-print
In this paper, we tackle the problem of enhancing real-world low-light images with significant noise in an unsupervised fashion.  ...  We propose to learn a two-stage GAN-based framework to enhance the real-world low-light images in a fully unsupervised fashion.  ...  Due to the difficulty of acquiring paired data in realworld scenarios, several weakly supervised and unsupervised enhancement approaches have been proposed, such a WESPE [37] , Deep Photo Enhancer [20  ... 
arXiv:2005.02818v2 fatcat:26npb5rnlvgv3deu3ypusi7nxq

Rain Removal and Illumination Enhancement Done in One Go [article]

Yecong Wan, Yuanshuo Cheng, Mingwen Shao
2021 arXiv   pre-print
In addition, after a simple transformation, our method outshines existing methods in both rain removal and low-light image enhancement.  ...  Therefore, it is very indispensable to jointly remove the rain and enhance the light for real-world rain image restoration. In this paper, we aim to address this problem from two aspects.  ...  [17] introduced an unsupervised GAN-based method that can be trained without low/normallight image pairs. Upon Retinex rule, Liu et al.  ... 
arXiv:2108.03873v2 fatcat:s3qatd3kajg3hmrnmblx7h4at4

A Switched View of Retinex: Deep Self-Regularized Low-Light Image Enhancement

Zhuqing Jiang, Haotian Li, Liangjie Liu, Aidong Men, Haiying Wang
2021 Neurocomputing  
A B S T R A C T Self-regularized low-light image enhancement does not require any normal-light image in training, thereby freeing from the chains of paired or unpaired training data that are time-consuming  ...  Our code is available at https://github.com/Github-LHT/A-Switched-View-of-Retinex-D eep-Self-Regularized-Low-Light-Image-Enhancement.  ...  Most of the existing deep learning-based methods are trained on paired or unpaired low-light and normal light images by supervised learning or generative adversarial network(GAN).  ... 
doi:10.1016/j.neucom.2021.05.025 fatcat:qfwoaf3r3fgrfg6gsjkqjeue2y

RetinexGAN:Unsupervised Low-light Enhancement with Two-layer Convolutional Decomposition Networks

Tian Ma, Ming Guo, Zhenhua Yu, Yanping Chen, Xincheng Ren, Runtao Xi, Yuancheng Li, Xinlei Zhou
2021 IEEE Access  
[31] proposes the DRBN (deep recursive band network) method, which uses paired low-light/normal-light images to restore and enhance the linear band representation of normallight images and then uses  ...  METHODS BASED ON DEEP LEARNING With the maturity of deep learning technology, it has been increasingly applied to the field of image enhancement, which has led to new developments in low-light image enhancement  ...  CONCLUSION In this paper, we propose a deep learning model based on generative adversarial networks and Retinex to enhance lowlight images with unpaired data-sets.  ... 
doi:10.1109/access.2021.3072331 fatcat:24s2jjbgizhaloi54pp3h6j3xa
« Previous Showing results 1 — 15 out of 52 results