Learning Raw Image Reconstruction-Aware Deep Image Compressors

Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG. Conventional and deep learning-based compression algorithms focus on minimizing image fidelity errors in the nonlinear standard RGB (sRGB) color space....

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on pattern analysis and machine intelligence Jg. 42; H. 4; S. 1013 - 1019
Hauptverfasser: Punnappurath, Abhijith, Brown, Michael S.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States IEEE 01.04.2020
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
ISSN:0162-8828, 1939-3539, 2160-9292, 1939-3539
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG. Conventional and deep learning-based compression algorithms focus on minimizing image fidelity errors in the nonlinear standard RGB (sRGB) color space. However, for many computer vision tasks, the sensor's linear raw-RGB image is desirable. Recent work has shown that the original raw-RGB image can be reconstructed using only small amounts of metadata embedded inside the JPEG image [1]. However, [1] relied on the conventional JPEG encoding that is unaware of the raw-RGB reconstruction task. In this paper, we examine the ability of deep image compressors to be "aware" of the additional objective of raw reconstruction. Towards this goal, we describe a general framework that enables deep networks targeting image compression to jointly consider both image fidelity errors and raw reconstruction errors. We describe this approach in two scenarios: (1) the network is trained from scratch using our proposed joint loss, and (2) a network originally trained only for sRGB fidelity loss is later fine-tuned to incorporate our raw reconstruction loss. When compared to sRGB fidelity-only compression, our combined loss leads to appreciable improvements in PSNR of the raw reconstruction with only minor impact on sRGB fidelity as measured by MS-SSIM.
AbstractList Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG. Conventional and deep learning-based compression algorithms focus on minimizing image fidelity errors in the nonlinear standard RGB (sRGB) color space. However, for many computer vision tasks, the sensor's linear raw-RGB image is desirable. Recent work has shown that the original raw-RGB image can be reconstructed using only small amounts of metadata embedded inside the JPEG image [1]. However, [1] relied on the conventional JPEG encoding that is unaware of the raw-RGB reconstruction task. In this paper, we examine the ability of deep image compressors to be "aware" of the additional objective of raw reconstruction. Towards this goal, we describe a general framework that enables deep networks targeting image compression to jointly consider both image fidelity errors and raw reconstruction errors. We describe this approach in two scenarios: (1) the network is trained from scratch using our proposed joint loss, and (2) a network originally trained only for sRGB fidelity loss is later fine-tuned to incorporate our raw reconstruction loss. When compared to sRGB fidelity-only compression, our combined loss leads to appreciable improvements in PSNR of the raw reconstruction with only minor impact on sRGB fidelity as measured by MS-SSIM.
Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG. Conventional and deep learning-based compression algorithms focus on minimizing image fidelity errors in the nonlinear standard RGB (sRGB) color space. However, for many computer vision tasks, the sensor's linear raw-RGB image is desirable. Recent work has shown that the original raw-RGB image can be reconstructed using only small amounts of metadata embedded inside the JPEG image [1]. However, [1] relied on the conventional JPEG encoding that is unaware of the raw-RGB reconstruction task. In this paper, we examine the ability of deep image compressors to be "aware" of the additional objective of raw reconstruction. Towards this goal, we describe a general framework that enables deep networks targeting image compression to jointly consider both image fidelity errors and raw reconstruction errors. We describe this approach in two scenarios: (1) the network is trained from scratch using our proposed joint loss, and (2) a network originally trained only for sRGB fidelity loss is later fine-tuned to incorporate our raw reconstruction loss. When compared to sRGB fidelity-only compression, our combined loss leads to appreciable improvements in PSNR of the raw reconstruction with only minor impact on sRGB fidelity as measured by MS-SSIM.Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG. Conventional and deep learning-based compression algorithms focus on minimizing image fidelity errors in the nonlinear standard RGB (sRGB) color space. However, for many computer vision tasks, the sensor's linear raw-RGB image is desirable. Recent work has shown that the original raw-RGB image can be reconstructed using only small amounts of metadata embedded inside the JPEG image [1]. However, [1] relied on the conventional JPEG encoding that is unaware of the raw-RGB reconstruction task. In this paper, we examine the ability of deep image compressors to be "aware" of the additional objective of raw reconstruction. Towards this goal, we describe a general framework that enables deep networks targeting image compression to jointly consider both image fidelity errors and raw reconstruction errors. We describe this approach in two scenarios: (1) the network is trained from scratch using our proposed joint loss, and (2) a network originally trained only for sRGB fidelity loss is later fine-tuned to incorporate our raw reconstruction loss. When compared to sRGB fidelity-only compression, our combined loss leads to appreciable improvements in PSNR of the raw reconstruction with only minor impact on sRGB fidelity as measured by MS-SSIM.
Author Punnappurath, Abhijith
Brown, Michael S.
Author_xml – sequence: 1
  givenname: Abhijith
  orcidid: 0000-0003-3438-5896
  surname: Punnappurath
  fullname: Punnappurath, Abhijith
  email: pabhijith@eecs.yorku.ca
  organization: Department of Electrical Engineering and Computer Science, York University, Toronto, ON, Canada
– sequence: 2
  givenname: Michael S.
  surname: Brown
  fullname: Brown, Michael S.
  email: mbrown@eecs.yorku.ca
  organization: Department of Electrical Engineering and Computer Science, York University, Toronto, ON, Canada
BackLink https://www.ncbi.nlm.nih.gov/pubmed/30843804$$D View this record in MEDLINE/PubMed
BookMark eNp9kU1LAzEQhoNU7If-AQUpePGydZJs0uxFKPWrUFFKPYeYnS1bupua7FL8925t66EHT3OY53kZ5u2SVulKJOSSwoBSSO7m76PXyYABTQYsAQ6SnZAOoxKihCWsRTpAJYuUYqpNuiEsAWgsgJ-RNgcVcwVxh9xP0fgyLxf9mdn0J4VZYH-G1pWh8rWtcldGo43x2H9AXO_3Y1esPYbgfDgnp5lZBbzYzx75eHqcj1-i6dvzZDyaRpYLWkUSUQnDJUjIBM3SOLMihYxKY6gVEGcpFyBEikMlh0Y2S8ubwYdJI6C1vEdud7lr775qDJUu8mBxtTIlujpoRpVKFIASDXpzhC5d7cvmOs34MJYJ5zxuqOs9VX8WmOq1zwvjv_XhMw2gdoD1LgSPmbZ5ZbYPqbzJV5qC3pagf0vQ2xL0voRGZUfqIf1f6Won5Yj4JygpFAXFfwCinpCR
CODEN ITPIDJ
CitedBy_id crossref_primary_10_32604_cmc_2020_014227
crossref_primary_10_1016_j_cviu_2022_103525
crossref_primary_10_3390_s24061743
crossref_primary_10_1109_TIM_2025_3544706
crossref_primary_10_1109_COMST_2023_3333342
crossref_primary_10_1109_TPAMI_2024_3356557
crossref_primary_10_1016_j_sigpro_2021_108346
crossref_primary_10_1109_TCSII_2021_3058245
crossref_primary_10_1016_j_fsidi_2021_301108
crossref_primary_10_1109_TCSVT_2023_3241319
crossref_primary_10_1109_TIP_2022_3214077
crossref_primary_10_3390_chemosensors10050164
crossref_primary_10_3390_e23121680
crossref_primary_10_1109_ACCESS_2023_3324403
crossref_primary_10_1007_s11263_024_02143_2
crossref_primary_10_1016_j_engappai_2023_106361
crossref_primary_10_1145_3592615
Cites_doi 10.1109/ICCV.2011.6126492
10.1007/978-3-642-33718-5_40
10.1109/CVPR.2016.183
10.1109/TPAMI.2003.1240119
10.1109/TPAMI.2012.58
10.1109/30.125072
10.1109/TPAMI.2014.2318713
10.1109/CVPR.2016.57
10.1145/1401132.1401174
10.5244/C.23.51
10.1109/CVPR.2018.00674
10.1109/ICCV.2015.42
10.1007/978-3-319-46448-0_26
10.1109/CVPR.2017.577
10.1109/ACSSC.2003.1292216
10.1109/CVPR.2018.00462
10.1109/ICCV.2017.190
10.1109/CVPR.1999.786966
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
DBID 97E
RIA
RIE
AAYXX
CITATION
NPM
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
DOI 10.1109/TPAMI.2019.2903062
DatabaseName IEEE All-Society Periodicals Package (ASPP) 2005–Present
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Xplore (IEEE/IET Electronic Library - IEL)
CrossRef
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 CrossRef
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
PubMed
Technology Research Database
MEDLINE - Academic
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: RIE
  name: IEEE Xplore (IEEE/IET Electronic Library - IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
Computer Science
EISSN 2160-9292
1939-3539
EndPage 1019
ExternalDocumentID 30843804
10_1109_TPAMI_2019_2903062
8658108
Genre orig-research
Journal Article
GrantInformation_xml – fundername: Canada First Research Excellence Fund
  funderid: 10.13039/501100010785
– fundername: Natural Sciences and Engineering Research Council of Canada; NSERC
  funderid: 10.13039/501100000038
GroupedDBID ---
-DZ
-~X
.DC
0R~
29I
4.4
53G
5GY
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACGFO
ACGFS
ACIWK
ACNCT
AENEX
AGQYO
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
ASUFR
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CS3
DU5
E.L
EBS
EJD
F5P
HZ~
IEDLZ
IFIPE
IPLJI
JAVBF
LAI
M43
MS~
O9-
OCL
P2P
PQQKQ
RIA
RIE
RNS
RXW
TAE
TN5
UHB
~02
AAYXX
CITATION
5VS
9M8
AAYOK
ABFSI
ADRHT
AETIX
AGSQL
AI.
AIBXA
ALLEH
FA8
H~9
IBMZZ
ICLAB
IFJZH
NPM
PKN
RIC
RIG
RNI
RZB
VH1
XJT
Z5M
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
ID FETCH-LOGICAL-c351t-6ee85a36060f51fd4fc5d0f16aa1c504fd35055de7867a6c5dc3a6c37960fecc3
IEDL.DBID RIE
ISICitedReferencesCount 29
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000526541100016&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 0162-8828
1939-3539
IngestDate Sun Sep 28 06:51:22 EDT 2025
Sun Nov 09 08:32:36 EST 2025
Wed Feb 19 02:30:52 EST 2025
Tue Nov 18 22:27:46 EST 2025
Sat Nov 29 05:15:58 EST 2025
Wed Aug 27 02:35:25 EDT 2025
IsPeerReviewed true
IsScholarly true
Issue 4
Language English
License https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html
https://doi.org/10.15223/policy-029
https://doi.org/10.15223/policy-037
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c351t-6ee85a36060f51fd4fc5d0f16aa1c504fd35055de7867a6c5dc3a6c37960fecc3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ORCID 0000-0003-3438-5896
PMID 30843804
PQID 2374693334
PQPubID 85458
PageCount 7
ParticipantIDs proquest_miscellaneous_2188980085
proquest_journals_2374693334
pubmed_primary_30843804
ieee_primary_8658108
crossref_citationtrail_10_1109_TPAMI_2019_2903062
crossref_primary_10_1109_TPAMI_2019_2903062
PublicationCentury 2000
PublicationDate 2020-04-01
PublicationDateYYYYMMDD 2020-04-01
PublicationDate_xml – month: 04
  year: 2020
  text: 2020-04-01
  day: 01
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
– name: New York
PublicationTitle IEEE transactions on pattern analysis and machine intelligence
PublicationTitleAbbrev TPAMI
PublicationTitleAlternate IEEE Trans Pattern Anal Mach Intell
PublicationYear 2020
Publisher IEEE
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: IEEE
– name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References ref31
ref30
ref2
ref1
ref19
(ref33) 0
theis (ref17) 2017
gregor (ref13) 2016
ballé (ref12) 2018
toderici (ref18) 2015; abs 1511 6085
(ref22) 0
kim (ref5) 2012; 34
ref24
rippel (ref16) 2017
ref26
ref25
ref20
kingma (ref32) 2014; abs 1412 6980
li (ref15) 2017; abs 1703 10553
ref28
mann (ref6) 1995
ref27
ref29
ref8
bellard (ref23) 0
ref7
agustsson (ref10) 2017
johnston (ref14) 2017; abs 1703 10114
ref9
ref4
ref3
taubman (ref21) 2001
ballé (ref11) 2017
References_xml – ident: ref26
  doi: 10.1109/ICCV.2011.6126492
– ident: ref27
  doi: 10.1007/978-3-642-33718-5_40
– start-page: 442
  year: 1995
  ident: ref6
  article-title: On being 'undigital' with digital cameras: Extending dynamic range by combining differently exposed pictures
  publication-title: Society for Imaging Science and Technology
– ident: ref1
  doi: 10.1109/CVPR.2016.183
– ident: ref4
  doi: 10.1109/TPAMI.2003.1240119
– volume: abs 1412 6980
  year: 2014
  ident: ref32
  article-title: Adam: A method for stochastic optimization
  publication-title: CoRR
– volume: 34
  start-page: 2289
  year: 2012
  ident: ref5
  article-title: A new in-camera imaging model for color computer vision and its application
  publication-title: IEEE Trans Pattern Anal Mach Intell
  doi: 10.1109/TPAMI.2012.58
– ident: ref9
  doi: 10.1109/30.125072
– start-page: 2922
  year: 2017
  ident: ref16
  article-title: Real-time adaptive image compression
  publication-title: Proc Int Conf Mach Learn
– year: 0
  ident: ref33
  article-title: Kodak lossless true color image suite (PhotoCD PCD0992).
– ident: ref2
  doi: 10.1109/TPAMI.2014.2318713
– volume: abs 1703 10553
  year: 2017
  ident: ref15
  article-title: Learning convolutional networks for content-weighted image compression
  publication-title: CoRR
– ident: ref28
  doi: 10.1109/CVPR.2016.57
– ident: ref3
  doi: 10.1145/1401132.1401174
– ident: ref25
  doi: 10.5244/C.23.51
– year: 0
  ident: ref22
  article-title: WebP image format
– year: 2017
  ident: ref11
  article-title: End-to-end optimized image compression
  publication-title: Proc Int Conf Learn Representations
– volume: abs 1511 6085
  year: 2015
  ident: ref18
  article-title: Variable rate image compression with recurrent neural networks
  publication-title: CoRR
– year: 0
  ident: ref23
  article-title: BPG image format
– ident: ref30
  doi: 10.1109/CVPR.2018.00674
– year: 2001
  ident: ref21
  publication-title: JPEG 2000 Image Compression Fundamentals Standards and Practice
– start-page: 1142
  year: 2017
  ident: ref10
  article-title: Soft-to-hard vector quantization for end-to-end learning compressible representations
  publication-title: Proc Conf Neural Inf Process Syst
– year: 2018
  ident: ref12
  article-title: Variational image compression with a scale hyperprior
  publication-title: Proc Int Conf Learn Representations
– ident: ref29
  doi: 10.1109/ICCV.2015.42
– ident: ref31
  doi: 10.1007/978-3-319-46448-0_26
– ident: ref19
  doi: 10.1109/CVPR.2017.577
– ident: ref20
  doi: 10.1109/ACSSC.2003.1292216
– volume: abs 1703 10114
  year: 2017
  ident: ref14
  article-title: Improved lossy image compression with priming and spatially adaptive bit rates for recurrent networks
  publication-title: CoRR
– ident: ref24
  doi: 10.1109/CVPR.2018.00462
– ident: ref8
  doi: 10.1109/ICCV.2017.190
– start-page: 3556
  year: 2016
  ident: ref13
  article-title: Towards conceptual compression
  publication-title: Proc Conf Neural Inf Process Syst
– ident: ref7
  doi: 10.1109/CVPR.1999.786966
– year: 2017
  ident: ref17
  article-title: Lossy image compression with compressive autoencoders
  publication-title: Proc Int Conf Learn Representations
SSID ssj0014503
Score 2.4630325
Snippet Deep learning-based image compressors are actively being explored in an effort to supersede conventional image compression algorithms, such as JPEG....
SourceID proquest
pubmed
crossref
ieee
SourceType Aggregation Database
Index Database
Enrichment Source
Publisher
StartPage 1013
SubjectTerms Accuracy
Algorithms
Calibration
Cameras
Compressors
Computer vision
Deep learning
deep learning-based image compression
Image coding
Image compression
Image reconstruction
JPEG encoders-decoders
Machine learning
radiometric calibration
raw image reconstruction
Table lookup
Transform coding
Title Learning Raw Image Reconstruction-Aware Deep Image Compressors
URI https://ieeexplore.ieee.org/document/8658108
https://www.ncbi.nlm.nih.gov/pubmed/30843804
https://www.proquest.com/docview/2374693334
https://www.proquest.com/docview/2188980085
Volume 42
WOSCitedRecordID wos000526541100016&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 Xplore (IEEE/IET Electronic Library - IEL)
  customDbUrl:
  eissn: 2160-9292
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0014503
  issn: 0162-8828
  databaseCode: RIE
  dateStart: 19790101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lb9QwEB6VigMcKLQ8AqUKEjdIG8dObF-QVkBFD1QVKtLeIj8mFRLsVrst_fvMeJ0AEiBxSiQ7ieWZsefLeOYDeCkG7IJDVTlpsVJGhcoYryqrvQjetNE3MZFN6NNTM5_bsy14PeXCIGI6fIaHfJti-XEZrvlX2ZGh7VJwZu8trbtNrtYUMVBtYkEmD4YsnGDEmCBT26Pzs9nHEz7FZQ8byz4yU9jI2nCxdfXbfpQIVv7ua6Y953jn_0Z7H-5l37KcbZThAWzhYhd2Rt6GMpvxLtz9pQjhHrzJJVYvyk_upjz5RgtMyaD0Z2nZanbjVli-Q7zM7fxOxunL1fohfD5-f_72Q5VZFaogW3FVdYimdZKASz20YohqCG2sB9E5J0JbqyFK8oraiNp02nXUGCRdpCasM5DA5SPYXiwX-ATKxnutOJAapVYq0hpuvfXoOTbYDHUoQIxz24dccpyZL772CXrUtk-i6Vk0fRZNAa-mZy43BTf-2XuPJ37qmee8gP1RhH22yXXf0CA7K6VUBbyYmsmaOETiFri8pj7CGGvYDy3g8Ub007tHjXn6528-gzsNY_F0qmcftklG-Bxuh-9XX9arA1LZuTlIKvsDEF7kew
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3fSxwxEB7EFmwfaqu13artCn1rVzeb7G7yIhxa8ageUq7g25IfsyLondxp_febyWW3Cm3Bp11INhsymWS-TGY-gM-sxcpqFJnmCjMhhc2kNCJTtWHWyNKZwgWyiXo0kufn6mwJvvaxMIgYLp_hLr0GX76b2js6KtuTfrtkFNn7jJizYrRW7zMQZeBB9jaM13EPJLoQmVztjc8Gp0O6x6V2C0VWMpHY8FxSunXxaEcKFCv_tjbDrnO0-rT-voZX0bpMB4vp8AaWcLIGqx1zQxoVeQ1ePkhDuA77McnqRfpD36fDa7_EpARL_ySXzQb3eobpIeJNLKc2CalPZ_O38PPo2_jgOIu8CpnlJbvNKkRZau6hS96WrHWitaXLW1ZpzWyZi9ZxbxeVDmtZ1bryhZb7B6892mm9yPkGLE-mE3wPaWFMLciV6ngthPOruDLKoCHvYNHmNgHWjW1jY9Jx4r64agL4yFUTRNOQaJoomgS-9N_cLFJu_Lf2Og18XzOOeQJbnQibqJXzpvCdrBTnXCSw0xd7fSIniZ7g9M7XYVIqSZZoAu8Wou_b7mbMh7__8xOsHI9PT5qT4ej7JrwoCJmHOz5bsOzlhdvw3P66vZzPPoaJ-xu-d-bc
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=Learning+Raw+Image+Reconstruction-Aware+Deep+Image+Compressors&rft.jtitle=IEEE+transactions+on+pattern+analysis+and+machine+intelligence&rft.au=Punnappurath%2C+Abhijith&rft.au=Brown%2C+Michael+S&rft.date=2020-04-01&rft.eissn=1939-3539&rft.volume=42&rft.issue=4&rft.spage=1013&rft_id=info:doi/10.1109%2FTPAMI.2019.2903062&rft_id=info%3Apmid%2F30843804&rft.externalDocID=30843804
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0162-8828&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0162-8828&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0162-8828&client=summon