NTIRE 2019 Image Dehazing Challenge Report

Ancuti, Cosmin; Timofte, Radu; Gool, Luc Van; Zhang, Lei; Yang, Ming–Hsuan; Guo, Tiantong; Li, Xuelu; Cherukuri, Venkateswararao; Monga, Vishal; Jiang, Hao; Yang, Siyuan; Liu, Yan; Qu, Xiaobo; Wan, Pengfei; Park, Dong-Won; Chun, Se Young; Hong, Minghui; Jin-ying, Huang; Chen, Yizi; Chen, Shuxin; Wang, Bomin; Michelini, Pablo Navarrete; Liu, Hanwen; Zhu, Dan; Jing, Liu; Santra, Sanchayan; Mondal, Ranjan; Chanda, Bhabatosh; Morales, Peter; Klinghoffer, Tzofi; Quan, Le Manh; Kim, Yong-Guk; Liang, Xiao; Li, Runde; Pan, Jinshan; Tang, Jinhui; Purohit, Kuldeep; Suin, Maitreya; Rajagopalan, A. N.; Schettini, Raimondo; Bianco, Simone; Piccoli, Flavio; Cusano, Claudio; Celona, Luigi; Hwang, Sunhee; Ma, Yu Seung; Byun, Hyeran; Murala, Subrahmanyam; Dudhane, Akshay; Harsh, Aulakh; Zheng, Tianxiang; Zhang, Tao; Weining, Qin; Zhou, Runnan; Shan-hu, Wang; Tarel, Jean-Philippe; Wang, Chuansheng; Wu, Jiawei

doi:10.1109/cvprw.2019.00277

Cited by 41 publications

(14 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The applications of the HRNet are not limited to the above that we have done, and are suitable to other positionsensitive vision applications, such as super-resolution, optical flow estimation, depth estimation, and so on. There are already followup works, e.g., image stylization [63], inpainting [35], image enhancement [46], image dehazing [1], and temporal pose estimation [4].…”

Section: Discussionmentioning

confidence: 99%

“…The main technical novelties compared with [105] lie in threefold. (1) We extend the network (named as HRNetV1) proposed in [105], to two versions: HRNetV2 and HRNetV2p, which explore all the four-resolution representations. (2) We build the connection between multi-resolution fusion and regular convolution, which provides an evidence for the necessity of exploring all the four-resolution representations in HRNetV2 and HRNetV2p.…”

Section: Multi-scale Fusionmentioning

confidence: 99%

See 1 more Smart Citation

Deep High-Resolution Representation Learning for Visual Recognition

Wang

Sun

Cheng

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

2,878

1,479

View full text Add to dashboard Cite

High-resolution representations are essential for position-sensitive vision problems, such as human pose estimation, semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions in series (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, our proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams in parallel; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise. We show the superiority of the proposed HRNet in a wide range of applications, including human pose estimation, semantic segmentation, and object detection, suggesting that the HRNet is a stronger backbone for computer vision problems. All the codes are available at https://github.com/HRNet. ! 1 INTRODUCTION D EEP convolutional neural networks (DCNNs) have achieved state-of-the-art results in many computer vision tasks, such as image classification, object detection, semantic segmentation, human pose estimation, and so on. The strength is that DCNNs are able to learn richer representations than conventional hand-crafted representations. Most recently-developed classification networks, including AlexNet [59], VGGNet [101], GoogleNet [108], ResNet [39], etc., follow the design rule of LeNet-5 [61]. This is depicted in Figure 1 (a): gradually reduce the spatial size of the feature maps, connect the convolutions from high resolution to low resolution in series, and lead to a low-resolution representation, which is further processed for classification.High-resolution representations are needed for positionsensitive tasks, e.g., semantic segmentation, human pose estimation, and object detection. The previous state-of-the-art methods adopt the high-resolution recovery process to raise the representation resolution from the low-resolution representation outputted by a classification or classification-like network as depicted in Figure 1 (b), e.g., Hourglass [83], Seg-Net [3], DeconvNet [85], U-Net [95], SimpleBaseline [124], and encoder-decoder [90]. In addition, dilated convolutions are used to remove some down-sample layers and thus yield medium-resolution representations [15], [144].We present a novel architecture, namely High-Resolution Net (HRNet), which is able to maintain high-resolution representations through the whole process. We start from a highresolution convolution stream, gradually add high-to-low resolution convolution streams one by one, and connect the multi-resolution streams in parallel. The resulting network • J. Wang is with Microsoft Research,

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Multi-scale Fusionmentioning

confidence: 99%

Deep High-Resolution Representation Learning for Visual Recognition

Wang

Sun

Cheng

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

2,878

1,479

View full text Add to dashboard Cite

show abstract

“…In addition, we implement our method on real-world dehazing benchmarks with the O-HAZE dataset [ 51 ] utilized in NTIRE2018 Dehazing Challenge [ 52 ], DENSE-HAZE dataset [ 53 ] utilized in NTIRE2019 Dehazing Challenge [ 54 ] and NH-HAZE dataset [ 55 ] utilized in the NTIRE2020 Dehazing Challenge [ 56 ]. O-HAZE, DENSE-HAZE and NH-HAZE contain 45 outdoor hazy images, 55 dense hazy images and 55 nonhomogeneous hazy images with their corresponding ground truth, respectively.…”

Section: Resultsmentioning

confidence: 99%

Residual Spatial and Channel Attention Networks for Single Image Dehazing

Jiang

Zhao

Zhu

et al. 2021

Sensors

View full text Add to dashboard Cite

Single image dehazing is a highly challenging ill-posed problem. Existing methods including both prior-based and learning-based heavily rely on the conceptual simplified atmospheric scattering model by estimating the so-called medium transmission map and atmospheric light. However, the formation of haze in the real world is much more complicated and inaccurate estimations further degrade the dehazing performance with color distortion, artifacts and insufficient haze removal. Moreover, most dehazing networks treat spatial-wise and channel-wise features equally, but haze is practically unevenly distributed across an image, thus regions with different haze concentrations require different attentions. To solve these problems, we propose an end-to-end trainable densely connected residual spatial and channel attention network based on the conditional generative adversarial framework to directly restore a haze-free image from an input hazy image, without explicitly estimation of any atmospheric scattering parameters. Specifically, a novel residual attention module is proposed by combining spatial attention and channel attention mechanism, which could adaptively recalibrate spatial-wise and channel-wise feature weights by considering interdependencies among spatial and channel information. Such a mechanism allows the network to concentrate on more useful pixels and channels. Meanwhile, the dense network can maximize the information flow along features from different levels to encourage feature reuse and strengthen feature propagation. In addition, the network is trained with a multi-loss function, in which contrastive loss and registration loss are novel refined to restore sharper structures and ensure better visual quality. Experimental results demonstrate that the proposed method achieves the state-of-the-art performance on both public synthetic datasets and real-world images with more visually pleasing dehazed results.

show abstract

“…As shown in Figure 17 (d)-(f), there still remains visible hazes in the dehazed image by our method. As a comparison, iPal-DH [64], trained on Dense Haze dataset [65], is able to dehaze images with dense haze more effectively, but it also fails to recover lost details covered by hazes.…”

Section: B Failed Casesmentioning

confidence: 99%

A Model-Driven Deep Dehazing Approach by Learning Deep Priors

Dong

Sun

2021

IEEE Access

View full text Add to dashboard Cite

Photos taken in hazy weather are usually covered with white masks and lose important details. Haze removal is a fundamental task and a prerequisite to many other vision tasks. Single image dehazing is an ill-posed inverse problem that has attracted much attention in recent years. Generally, current single dehazing methods can be categorized into the traditional prior-based methods and the data-driven deep learning methods that respectively investigate haze-related image priors and deep architectures. In this paper, we propose a novel model-driven deep learning approach that combines the advantages of both kinds of methods. First, we build an energy model for single image dehazing with physical constraints in both color image space and haze-related feature space (implemented as dark channel space in this work), regularized by haze-related image priors. Then, we design an iterative optimization algorithm for solving the proposed dehazing energy model based on the half-quadratic splitting algorithm, and the priors are transformed to their corresponding proximal operators. Finally, inspired by the optimization algorithm, we design a deep dehazing neural network, dubbed as proximal dehaze-net, by learning the proximal operators for haze-related image priors using CNNs. Our network incorporates physical model constraints of hazes and haze-related prior learning into a novel deep architecture. Extensive experiments show that our method achieves promising performance for single image dehazing.

show abstract

NTIRE 2019 Image Dehazing Challenge Report

Cited by 41 publications

References 29 publications

Deep High-Resolution Representation Learning for Visual Recognition

Deep High-Resolution Representation Learning for Visual Recognition

Residual Spatial and Channel Attention Networks for Single Image Dehazing

A Model-Driven Deep Dehazing Approach by Learning Deep Priors

Contact Info

Product

Resources

About