Single image dehazing via multi-scale convolutional neural networks

Wenqi Ren, Si Liu, Hua Zhang, Jinshan Pan, Xiaochun Cao, Ming Hsuan Yang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

222 Citations (Scopus)

Abstract

The performance of existing image dehazing methods is limited by hand-designed features, such as the dark channel, color disparity and maximum contrast, with complex fusion schemes. In this paper, we propose a multi-scale deep neural network for single-image dehazing by learning the mapping between hazy images and their corresponding transmission maps. The proposed algorithm consists of a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale net which refines results locally. To train the multiscale deep network, we synthesize a dataset comprised of hazy images and corresponding transmission maps based on the NYU Depth dataset. Extensive experiments demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods on both synthetic and real-world images in terms of quality and speed.

Original languageEnglish
Title of host publicationComputer Vision - 14th European Conference, ECCV 2016, Proceedings
EditorsBastian Leibe, Nicu Sebe, Max Welling, Jiri Matas
PublisherSpringer Verlag
Pages154-169
Number of pages16
ISBN (Print)9783319464749
DOIs
Publication statusPublished - 2016 Jan 1
Event14th European Conference on Computer Vision, ECCV 2016 - Amsterdam, Netherlands
Duration: 2016 Oct 82016 Oct 16

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9906 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th European Conference on Computer Vision, ECCV 2016
CountryNetherlands
CityAmsterdam
Period16/10/816/10/16

Fingerprint

Neural Networks
Neural networks
Fusion reactions
Color
Fusion
Experiments
Entire
Predict
Demonstrate
Experiment
Deep neural networks

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ren, W., Liu, S., Zhang, H., Pan, J., Cao, X., & Yang, M. H. (2016). Single image dehazing via multi-scale convolutional neural networks. In B. Leibe, N. Sebe, M. Welling, & J. Matas (Eds.), Computer Vision - 14th European Conference, ECCV 2016, Proceedings (pp. 154-169). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9906 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46475-6_10
Ren, Wenqi ; Liu, Si ; Zhang, Hua ; Pan, Jinshan ; Cao, Xiaochun ; Yang, Ming Hsuan. / Single image dehazing via multi-scale convolutional neural networks. Computer Vision - 14th European Conference, ECCV 2016, Proceedings. editor / Bastian Leibe ; Nicu Sebe ; Max Welling ; Jiri Matas. Springer Verlag, 2016. pp. 154-169 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{38c5d15b37734c23ac50fa6bf346a153,
title = "Single image dehazing via multi-scale convolutional neural networks",
abstract = "The performance of existing image dehazing methods is limited by hand-designed features, such as the dark channel, color disparity and maximum contrast, with complex fusion schemes. In this paper, we propose a multi-scale deep neural network for single-image dehazing by learning the mapping between hazy images and their corresponding transmission maps. The proposed algorithm consists of a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale net which refines results locally. To train the multiscale deep network, we synthesize a dataset comprised of hazy images and corresponding transmission maps based on the NYU Depth dataset. Extensive experiments demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods on both synthetic and real-world images in terms of quality and speed.",
author = "Wenqi Ren and Si Liu and Hua Zhang and Jinshan Pan and Xiaochun Cao and Yang, {Ming Hsuan}",
year = "2016",
month = "1",
day = "1",
doi = "10.1007/978-3-319-46475-6_10",
language = "English",
isbn = "9783319464749",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "154--169",
editor = "Bastian Leibe and Nicu Sebe and Max Welling and Jiri Matas",
booktitle = "Computer Vision - 14th European Conference, ECCV 2016, Proceedings",
address = "Germany",

}

Ren, W, Liu, S, Zhang, H, Pan, J, Cao, X & Yang, MH 2016, Single image dehazing via multi-scale convolutional neural networks. in B Leibe, N Sebe, M Welling & J Matas (eds), Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9906 LNCS, Springer Verlag, pp. 154-169, 14th European Conference on Computer Vision, ECCV 2016, Amsterdam, Netherlands, 16/10/8. https://doi.org/10.1007/978-3-319-46475-6_10

Single image dehazing via multi-scale convolutional neural networks. / Ren, Wenqi; Liu, Si; Zhang, Hua; Pan, Jinshan; Cao, Xiaochun; Yang, Ming Hsuan.

Computer Vision - 14th European Conference, ECCV 2016, Proceedings. ed. / Bastian Leibe; Nicu Sebe; Max Welling; Jiri Matas. Springer Verlag, 2016. p. 154-169 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9906 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Single image dehazing via multi-scale convolutional neural networks

AU - Ren, Wenqi

AU - Liu, Si

AU - Zhang, Hua

AU - Pan, Jinshan

AU - Cao, Xiaochun

AU - Yang, Ming Hsuan

PY - 2016/1/1

Y1 - 2016/1/1

N2 - The performance of existing image dehazing methods is limited by hand-designed features, such as the dark channel, color disparity and maximum contrast, with complex fusion schemes. In this paper, we propose a multi-scale deep neural network for single-image dehazing by learning the mapping between hazy images and their corresponding transmission maps. The proposed algorithm consists of a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale net which refines results locally. To train the multiscale deep network, we synthesize a dataset comprised of hazy images and corresponding transmission maps based on the NYU Depth dataset. Extensive experiments demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods on both synthetic and real-world images in terms of quality and speed.

AB - The performance of existing image dehazing methods is limited by hand-designed features, such as the dark channel, color disparity and maximum contrast, with complex fusion schemes. In this paper, we propose a multi-scale deep neural network for single-image dehazing by learning the mapping between hazy images and their corresponding transmission maps. The proposed algorithm consists of a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale net which refines results locally. To train the multiscale deep network, we synthesize a dataset comprised of hazy images and corresponding transmission maps based on the NYU Depth dataset. Extensive experiments demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods on both synthetic and real-world images in terms of quality and speed.

UR - http://www.scopus.com/inward/record.url?scp=84990840616&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990840616&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46475-6_10

DO - 10.1007/978-3-319-46475-6_10

M3 - Conference contribution

AN - SCOPUS:84990840616

SN - 9783319464749

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 154

EP - 169

BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings

A2 - Leibe, Bastian

A2 - Sebe, Nicu

A2 - Welling, Max

A2 - Matas, Jiri

PB - Springer Verlag

ER -

Ren W, Liu S, Zhang H, Pan J, Cao X, Yang MH. Single image dehazing via multi-scale convolutional neural networks. In Leibe B, Sebe N, Welling M, Matas J, editors, Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Springer Verlag. 2016. p. 154-169. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46475-6_10