SNR-Aware Low-light Image Enhancement

Xu, Xiaogang; Wang, Ruixing; Jia, Jiaya

doi:10.1109/cvpr52688.2022.01719

Cited by 175 publications

(52 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare our results with seven methods, including SRIE [16], LIME [14], EnlightenGAN [29], Zero-DCE [30], URetinex [36], SNRANet [27], and UHDFour [47], on LOL-V2 dataset. As shown in Table III, we outperform other comparison methods in PSNR and achieve the second-best performance in SSIM.…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

“…AGLLNet [22] and DRBN [23] integrated multiscale feature bands to enhance low-light images. To seek higher performance, various Transformer frameworks [27], [28], [44], invertible network [45], and normalizing flow [29] were embedded into enhancement models. DCCNet [46] explored the color consistency of low-light images via a color histogram.…”

Section: A Low-light Image Enhancementmentioning

confidence: 99%

“…Fueled by convolutional neural network, a large group of researchers have attempted to train high-efficient low-light image enhancement models from a sea of data. Deep blackbox methods directly learn a powerful mapping relationship from low-light images to normal-light images via denoising autoencoder [21], multi-scale framework [22]- [25], normalizing flow [26], transformer [27], [28], deep unsupervised framework [29], [30], and so on. However, they focus more on enhancement performance at the expense of the interpretability and physical degradation principle, which could heavily hinder the performance improvements.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Color Consistent Network for Low-Light Image Enhancement

Zhang¹,

Hong²,

Yan³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Section: A Low-light Image Enhancementmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Color Consistent Network for Low-Light Image Enhancement

Zhang¹,

Hong²,

Yan³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Images captured in low-light environments usually contain large scales of dark areas. They are of low contrast and intensities, which submerge the useful image contents, making them invisible to humans as well as damaging the performances of numerous computer vision algorithms * Equal contribution † Corresponding author Input SCI [28] URetinexNet [41] SNR [43] Ours GT Figure 1. This figure shows a challenging low-light image, whose middle areas are completely swallowed by the dark.…”

Section: Introductionmentioning

confidence: 99%

“…Transformers are introduced to enhance low-light images by modelling longer-range context dependencies [52,43]. An effective paradigm is to combine self-attention and convolution in a complementary way.…”

Section: Introductionmentioning

confidence: 99%

Low illumination image enhancement technology based on iterative dense residual network

Wang

Piao

Wang

et al. 2021

2021 International Conference on Electronic Information Engineering and Computer Science (EIECS)

View full text Add to dashboard Cite

This paper presents a novel network structure with illumination-aware gamma correction and complete image modelling to solve the low-light image enhancement problem. Low-light environments usually lead to less informative large-scale dark areas, directly learning deep representations from low-light images is insensitive to recovering normal illumination. We propose to integrate the effectiveness of gamma correction with the strong modelling capacities of deep networks, which enables the correction factor gamma to be learned in a coarse to elaborate manner via adaptively perceiving the deviated illumination. Because exponential operation introduces high computational complexity, we propose to use Taylor Series to approximate gamma correction, accelerating the training and inference speed. Dark areas usually occupy large scales in low-light images, common local modelling structures, e.g., CNN, SwinIR, are thus insufficient to recover accurate illumination across whole low-light images. We propose a novel Transformer block to completely simulate the dependencies of all pixels across images via a local-to-global hierarchical attention mechanism, so that dark areas could be inferred by borrowing the information from far informative regions in a highly effective manner. Extensive experiments on several benchmark datasets demonstrate that our approach outperforms state-of-the-art methods.

show abstract

CNSNet: A Cleanness-Navigated-Shadow Network for Shadow Removal

Zheng

Huang

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

SNR-Aware Low-light Image Enhancement

Cited by 175 publications

References 42 publications

Deep Color Consistent Network for Low-Light Image Enhancement

Deep Color Consistent Network for Low-Light Image Enhancement

Low illumination image enhancement technology based on iterative dense residual network

CNSNet: A Cleanness-Navigated-Shadow Network for Shadow Removal

Contact Info

Product

Resources

About