DEA-Net: Single Image Dehazing Based on Detail-Enhanced Convolution and Content-Guided Attention

Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE transactions on image processing Ročník 33; s. 1002
Hlavní autori: Chen, Zixuan, He, Zewei, Lu, Zhe-Ming
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: United States The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024
Predmet:
ISSN:1057-7149, 1941-0042, 1941-0042
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Abstract Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional Neural Network (CNN) structure is still under-explored. In this paper, a Detail-Enhanced Attention Block (DEAB) consisting of Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA) is proposed to boost the feature learning for improving the dehazing performance. Specifically, the DEConv contains difference convolutions which can integrate prior information to complement the vanilla one and enhance the representation capacity. Then by using the re-parameterization technique, DEConv is equivalently converted into a vanilla convolution to reduce parameters and computational cost. By assigning the unique Spatial Importance Map (SIM) to every channel, CGA can attend more useful information encoded in features. In addition, a CGA-based mixup fusion scheme is presented to effectively fuse the features and aid the gradient flow. By combining above mentioned components, we propose our Detail-Enhanced Attention Network (DEA-Net) for recovering high-quality haze-free images. Extensive experimental results demonstrate the effectiveness of our DEA-Net, outperforming the state-of-the-art (SOTA) methods by boosting the PSNR index over 41 dB with only 3.653 M parameters. (The source code of our DEA-Net is available at https://github.com/cecret3350/DEA-Net.).
AbstractList Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional Neural Network (CNN) structure is still under-explored. In this paper, a Detail-Enhanced Attention Block (DEAB) consisting of Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA) is proposed to boost the feature learning for improving the dehazing performance. Specifically, the DEConv contains difference convolutions which can integrate prior information to complement the vanilla one and enhance the representation capacity. Then by using the re-parameterization technique, DEConv is equivalently converted into a vanilla convolution to reduce parameters and computational cost. By assigning the unique Spatial Importance Map (SIM) to every channel, CGA can attend more useful information encoded in features. In addition, a CGA-based mixup fusion scheme is presented to effectively fuse the features and aid the gradient flow. By combining above mentioned components, we propose our Detail-Enhanced Attention Network (DEA-Net) for recovering high-quality haze-free images. Extensive experimental results demonstrate the effectiveness of our DEA-Net, outperforming the state-of-the-art (SOTA) methods by boosting the PSNR index over 41 dB with only 3.653 M parameters. (The source code of our DEA-Net is available at https://github.com/cecret3350/DEA-Net .)
Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional Neural Network (CNN) structure is still under-explored. In this paper, a Detail-Enhanced Attention Block (DEAB) consisting of Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA) is proposed to boost the feature learning for improving the dehazing performance. Specifically, the DEConv contains difference convolutions which can integrate prior information to complement the vanilla one and enhance the representation capacity. Then by using the re-parameterization technique, DEConv is equivalently converted into a vanilla convolution to reduce parameters and computational cost. By assigning the unique Spatial Importance Map (SIM) to every channel, CGA can attend more useful information encoded in features. In addition, a CGA-based mixup fusion scheme is presented to effectively fuse the features and aid the gradient flow. By combining above mentioned components, we propose our Detail-Enhanced Attention Network (DEA-Net) for recovering high-quality haze-free images. Extensive experimental results demonstrate the effectiveness of our DEA-Net, outperforming the state-of-the-art (SOTA) methods by boosting the PSNR index over 41 dB with only 3.653 M parameters. (The source code of our DEA-Net is available at https://github.com/cecret3350/DEA-Net.).Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional Neural Network (CNN) structure is still under-explored. In this paper, a Detail-Enhanced Attention Block (DEAB) consisting of Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA) is proposed to boost the feature learning for improving the dehazing performance. Specifically, the DEConv contains difference convolutions which can integrate prior information to complement the vanilla one and enhance the representation capacity. Then by using the re-parameterization technique, DEConv is equivalently converted into a vanilla convolution to reduce parameters and computational cost. By assigning the unique Spatial Importance Map (SIM) to every channel, CGA can attend more useful information encoded in features. In addition, a CGA-based mixup fusion scheme is presented to effectively fuse the features and aid the gradient flow. By combining above mentioned components, we propose our Detail-Enhanced Attention Network (DEA-Net) for recovering high-quality haze-free images. Extensive experimental results demonstrate the effectiveness of our DEA-Net, outperforming the state-of-the-art (SOTA) methods by boosting the PSNR index over 41 dB with only 3.653 M parameters. (The source code of our DEA-Net is available at https://github.com/cecret3350/DEA-Net.).
Author He, Zewei
Chen, Zixuan
Lu, Zhe-Ming
Author_xml – sequence: 1
  givenname: Zixuan
  surname: Chen
  fullname: Chen, Zixuan
– sequence: 2
  givenname: Zewei
  surname: He
  fullname: He, Zewei
– sequence: 3
  givenname: Zhe-Ming
  surname: Lu
  fullname: Lu, Zhe-Ming
BackLink https://www.ncbi.nlm.nih.gov/pubmed/38252568$$D View this record in MEDLINE/PubMed
BookMark eNpdkEtPwzAQhC0Eog-4c0KRuHBx8Ts2t9KWUqkCJMo5OI7TpkqckjhI8OtxoVw47c7Mp9VoB-DY1c4CcIHRCGOkblaL5xFBhI0o5QwjeQT6WDEMEWLkOOyIxzDGTPXAoG23CGHGsTgFPSoJJ1zIPnibzsbw0frb6KVw69JGi0qvbTS1G_0VjOhOtzaLahccr4sSztxGOxOsSe0-6rLzRci0-9HeOg_nXZGFeOz3KoRn4CTXZWvPD3MIXu9nq8kDXD7NF5PxEu6wUB7yFJtMCJ0bnGpGhUIGxTS3sWQ5kVqjnLNUh9LSqBTnwiArsjjLTcwYUUTRIbj-vbtr6vfOtj6pitbYstTO1l2bEIVjKQTDe_TqH7qtu8aFdntKCiYoJ4G6PFBdWtks2TVFpZvP5O939Bs3R3EZ
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024
DBID NPM
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
DOI 10.1109/TIP.2024.3354108
DatabaseName PubMed
Computer and Information Systems Abstracts
Electronics & Communications Abstracts
Technology Research Database
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
DatabaseTitle PubMed
Technology Research Database
Computer and Information Systems Abstracts – Academic
Electronics & Communications Abstracts
ProQuest Computer Science Collection
Computer and Information Systems Abstracts
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
DatabaseTitleList Technology Research Database
MEDLINE - Academic
PubMed
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Applied Sciences
Engineering
EISSN 1941-0042
ExternalDocumentID 38252568
Genre Journal Article
GroupedDBID ---
-~X
.DC
0R~
29I
4.4
53G
5GY
5VS
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABFSI
ABQJQ
ABVLG
ACGFO
ACGFS
ACIWK
AENEX
AETIX
AGQYO
AGSQL
AHBIQ
AI.
AIBXA
AKJIK
AKQYR
ALLEH
ALMA_UNASSIGNED_HOLDINGS
ASUFR
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CS3
DU5
E.L
EBS
EJD
F5P
HZ~
H~9
ICLAB
IFIPE
IFJZH
IPLJI
JAVBF
LAI
M43
MS~
NPM
O9-
OCL
P2P
RIA
RIE
RIG
RNS
TAE
TN5
VH1
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
ID FETCH-LOGICAL-p169t-5b1cd66afc1ba43690c073fe784f28aa0f54ba5258c9b1f6c0e6d7dfc74429293
ISICitedReferencesCount 364
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001166881600002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1057-7149
1941-0042
IngestDate Sat Sep 27 19:35:11 EDT 2025
Mon Jun 30 10:10:49 EDT 2025
Mon Jul 21 05:58:57 EDT 2025
IsPeerReviewed true
IsScholarly true
Language English
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-p169t-5b1cd66afc1ba43690c073fe784f28aa0f54ba5258c9b1f6c0e6d7dfc74429293
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
PMID 38252568
PQID 2918646352
PQPubID 85429
ParticipantIDs proquest_miscellaneous_2917866419
proquest_journals_2918646352
pubmed_primary_38252568
PublicationCentury 2000
PublicationDate 2024-00-00
20240101
PublicationDateYYYYMMDD 2024-01-01
PublicationDate_xml – year: 2024
  text: 2024-00-00
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
– name: New York
PublicationTitle IEEE transactions on image processing
PublicationTitleAlternate IEEE Trans Image Process
PublicationYear 2024
Publisher The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
SSID ssj0014516
Score 2.7435539
Snippet Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based...
SourceID proquest
pubmed
SourceType Aggregation Database
Index Database
StartPage 1002
SubjectTerms Artificial neural networks
Deep learning
Gradient flow
Haze
Ill posed problems
Image enhancement
Image quality
Machine learning
Parameterization
Parameters
Source code
Title DEA-Net: Single Image Dehazing Based on Detail-Enhanced Convolution and Content-Guided Attention
URI https://www.ncbi.nlm.nih.gov/pubmed/38252568
https://www.proquest.com/docview/2918646352
https://www.proquest.com/docview/2917866419
Volume 33
WOSCitedRecordID wos001166881600002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE Electronic Library (IEL)
  customDbUrl:
  eissn: 1941-0042
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0014516
  issn: 1057-7149
  databaseCode: RIE
  dateStart: 19920101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwtV3db9MwELfYQGg88DFgK4zJSLxNLnGT-IO3shY2CapJFFTxUhLHoZGYG63p6J_P2U6yMDEJHuhDVNmJ0-b3y53Pd75D6BVVIkyEe7-pIhETMZFM5sR6cSjVQZa4dcgvH_hkImYzeVYHsa9cOQFujNhsZPlfoYY2ANtunf0HuNtBoQG-A-hwBNjh-FfAj8ZDMtFuxe8T6KUf-uj03AbmjPTCppL-fvQWFFdmnQQjFz5KxmbhwwCOl-ay_mnOp-AyV5mKvF8XGXQPq8rHRnYntNZYtHUmmqLjzvtQuBuWfg9CoxtdDIEXcl-LzfqKlSfeQ6J_6qIND1q7poUmH5vL65UJvxW6r70clRElViB0BW0YdiSlTf36ZxHuMqBOT8_6dsx-GMYRdakfqg545blDLwT7FuZs4kqZtSGGTdcWuj3gsRR-a1_rYLL1iRuvdSBfX7_dDrrbDHCz6eGmINOH6H5tO-Chx_wRuqXNLnpQ2xG4ltKrXXSvk2TyMfpWE-IN9nTAjg64oQN2dMBLg6_RAXfogIEO-Hc64JYOT9Dnd-Pp8Qmp62qQkjJZkTilKmMsyRVNkyhkMlAg6HPNRZQPRJIEeRylCfx5oWRKc6YCzTKe5YpHtriZDJ-ibbM0eh_hTLA0DSJrBShQBblUUsYJfLIo54rTHjpoHt68fkdW84GkgkUw0x300Mu2G8Sa9VUlRi_X7hwuGIORe2jPP_R56fOvzBtknt3Y8xztWDT9UtkB2q4u1voFuqMuq2J1cYi2-EwcOj78AgSPcGE
linkProvider IEEE
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=DEA-Net%3A+Single+Image+Dehazing+Based+on+Detail-Enhanced+Convolution+and+Content-Guided+Attention&rft.jtitle=IEEE+transactions+on+image+processing&rft.au=Chen%2C+Zixuan&rft.au=He%2C+Zewei&rft.au=Lu%2C+Zhe-Ming&rft.date=2024&rft.eissn=1941-0042&rft.volume=33&rft.spage=1002&rft_id=info:doi/10.1109%2FTIP.2024.3354108&rft_id=info%3Apmid%2F38252568&rft.externalDocID=38252568
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1057-7149&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1057-7149&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1057-7149&client=summon