EnlightenGAN: Deep Light Enhancement without Paired Supervision

Jiang, Yifan; Gong, Xinyu; Liu, Ding; Chen, Fang; Shen, Xiaohui; Yang, Jianchao; Zhou, Pan; Wang, Zhangyang

doi:10.48550/arxiv.1906.06972

Cited by 47 publications

(118 citation statements)

References 58 publications

Supporting

Mentioning

118

Contrasting

Order By: Relevance

“…As Fig. 8 shows, our fused output brings back the missing details and appears more attractive than those generated by recent works [19,16].…”

Section: Applicationsmentioning

confidence: 86%

See 1 more Smart Citation

Neural Camera Simulators

Ouyang¹,

Shi²,

Lei³

et al. 2021

Preprint

View full text Add to dashboard Cite

We present a controllable camera simulator based on deep neural networks to synthesize raw image data under different camera settings, including exposure time, ISO, and aperture. The proposed simulator includes an exposure module that utilizes the principle of modern lens designs for correcting the luminance level. It also contains a noise module using the noise level function and an aperture module with adaptive attention to simulate the side effects on noise and defocus blur. To facilitate the learning of a simulator model, we collect a dataset of the 10,000 raw images of 450 scenes with different exposure settings. Quantitative experiments and qualitative comparisons show that our approach outperforms relevant baselines in raw data synthesize on multiple cameras. Furthermore, the camera simulator enables various applications, including large-aperture enhancement, HDR, auto exposure, and data augmentation for training local feature detectors. Our work represents the first attempt to simulate a camera sensor's behavior leveraging both the advantage of traditional raw sensor features and the power of data-driven deep learning.

show abstract

“…As Fig. 8 shows, our fused output brings back the missing details and appears more attractive than those generated by recent works [19,16].…”

Section: Applicationsmentioning

confidence: 86%

“…We pick two representative simulated images for display. Results from EnlightenGAN [19] and Zero-DCE [16] are listed in the last two rows. Zoom-in to view details.…”

Section: Applicationsmentioning

confidence: 99%

Neural Camera Simulators

Ouyang¹,

Shi²,

Lei³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…This kinds of methods adopt CNN to extract features from input images of initial size and reconstruct every pixel from dense pixel-to-pixel mapping or transformation operations. This kind of approaches have made great breakthroughs and achieved SOTA performance in many image enhancment tasks [11,22,29,3,24,16,2]. [10] proposes a residual CNN architecture as enhancer to learn the pixel-wise translation function between lowquality cellphone images and high-quality Digital Single-Lens Reflex (DSLR) images.…”

Section: Related Workmentioning

confidence: 99%

“…[10] proposes a residual CNN architecture as enhancer to learn the pixel-wise translation function between lowquality cellphone images and high-quality Digital Single-Lens Reflex (DSLR) images. [3,11,2,8] all employ UNetstyle structure originated from [18] for different image quality enhancement tasks. Despite their SOTA performance, these dense pixel-wise feature extraction and regeneration methods are too heavy to be used for practical applications, especially for high resolution input images [25].…”

Section: Related Workmentioning

confidence: 99%

Real-time Image Enhancer via Learnable Spatial-aware 3D Lookup Tables

Wang¹,

Li²,

Peng³

et al. 2021

Preprint

View full text Add to dashboard Cite

Recently, deep learning-based image enhancement algorithms achieved state-of-the-art (SOTA) performance on several publicly available datasets. However, most existing methods fail to meet practical requirements either for visual perception or for computation efficiency, especially for high-resolution images. In this paper, we propose a novel real-time image enhancer via learnable spatial-aware 3dimentional lookup tables(3D LUTs), which well considers global scenario and local spatial information. Specifically, we introduce a light weight two-head weight predictor that has two outputs. One is a 1D weight vector used for imagelevel scenario adaptation, the other is a 3D weight map aimed for pixel-wise category fusion. We learn the spatialaware 3D LUTs and fuse them according to the aforementioned weights in an end-to-end manner. The fused LUT is then used to transform the source image into the target tone in an efficient way. Extensive results show that our model outperforms SOTA image enhancement methods on public datasets both subjectively and objectively, and that our model only takes about 4ms to process a 4K resolution image on one NVIDIA V100 GPU.

show abstract

“…Following that, various other neural networks [22,27] have been proposed for supervised LLIE. More recent methods [8] focus on unsupervised LLIE which directly enlightens low-light images without any paired training data. The very recent Zero-DCE [5] trains the deep LLIE model using non-reference losses.…”

Section: Low-light Image Enhancementmentioning

confidence: 99%

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

Zhang¹,

Huang²,

Wen³

2021

Preprint

View full text Add to dashboard Cite

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 according to diverse preference by each individual. To tackle these two challenges, this paper presents a novel deep reinforcement learning based method, dubbed ReLLIE, for customized low-light enhancement. ReLLIE models LLIE as a markov decision process, i.e., estimating the pixel-wise image-specific curves sequentially and recurrently. Given the reward computed from a set of carefully crafted non-reference loss functions, a lightweight network is proposed to estimate the curves for enlightening a low-light image input. As ReLLIE learns a policy instead of one-one image translation, it can handle various low-light measurements and provide customized enhanced outputs by flexibly applying the policy different times. Furthermore, ReLLIE can enhance real-world images with hybrid corruptions, e.g., noise, by using a plug-and-play denoiser easily. Extensive experiments on various benchmarks demonstrate the advantages of ReLLIE, comparing to the state-of-the-art methods. (Code is available: https://github.com/GuoLanqing/ReLLIE.) CCS CONCEPTS• Computing methodologies → Image processing; Sequential decision making.

show abstract

EnlightenGAN: Deep Light Enhancement without Paired Supervision

Abstract: Figure 1: Representative visual examples by enhancing low-light images using EnlightenGAN.

Cited by 47 publications

References 58 publications

Neural Camera Simulators

Neural Camera Simulators

Real-time Image Enhancer via Learnable Spatial-aware 3D Lookup Tables

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

Contact Info

Product

Resources

About