Compact Representation of High-Dimensional Feature Vectors for Large-Scale Image Recognition and Retrieval

In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors (VLAD), have achieved state-of-the-art results. However, the combination of the large numbers of examples and high-dimensional vectors necessitates...

Full description

Saved in:

Bibliographic Details
Published in:	IEEE transactions on image processing Vol. 25; no. 5; pp. 2407 - 2419
Main Authors:	Zhang, Yu, Wu, Jianxin, Cai, Jianfei
Format:	Journal Article
Language:	English
Published:	United States IEEE 01.05.2016 The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:	Correlation Digital imaging feature selection Image coding Image recognition image representation Image retrieval large scale Quantization (signal) Sorting Visualization large scale Feature selection image representation
ISSN:	1057-7149, 1941-0042, 1941-0042
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Abstract	In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors (VLAD), have achieved state-of-the-art results. However, the combination of the large numbers of examples and high-dimensional vectors necessitates dimensionality reduction, in order to reduce its storage and CPU costs to a reasonable range. In spite of the popularity of various feature compression methods, this paper shows that the feature (dimension) selection is a better choice for high-dimensional FV/VLAD than the feature (dimension) compression methods, e.g., product quantization. We show that strong correlation among the feature dimensions in the FV and the VLAD may not exist, which renders feature selection a natural choice. We also show that, many dimensions in FV/VLAD are noise. Throwing them away using feature selection is better than compressing them and useful dimensions altogether using feature compression methods. To choose features, we propose an efficient importance sorting algorithm considering both the supervised and unsupervised cases, for visual recognition and image retrieval, respectively. Combining with the 1-bit quantization, feature selection has achieved both higher accuracy and less computational cost than feature compression methods, such as product quantization, on the FV and the VLAD image representations.
AbstractList	In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors (VLAD), have achieved state-of-the-art results. However, the combination of the large numbers of examples and high-dimensional vectors necessitates dimensionality reduction, in order to reduce its storage and CPU costs to a reasonable range. In spite of the popularity of various feature compression methods, this paper shows that the feature (dimension) selection is a better choice for high-dimensional FV/VLAD than the feature (dimension) compression methods, e.g., product quantization. We show that strong correlation among the feature dimensions in the FV and the VLAD may not exist, which renders feature selection a natural choice. We also show that, many dimensions in FV/VLAD are noise. Throwing them away using feature selection is better than compressing them and useful dimensions altogether using feature compression methods. To choose features, we propose an efficient importance sorting algorithm considering both the supervised and unsupervised cases, for visual recognition and image retrieval, respectively. Combining with the 1-bit quantization, feature selection has achieved both higher accuracy and less computational cost than feature compression methods, such as product quantization, on the FV and the VLAD image representations. In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors (VLAD), have achieved state-of-the-art results. However, the combination of the large numbers of examples and high-dimensional vectors necessitates dimensionality reduction, in order to reduce its storage and CPU costs to a reasonable range. In spite of the popularity of various feature compression methods, this paper shows that the feature (dimension) selection is a better choice for high-dimensional FV/VLAD than the feature (dimension) compression methods, e.g., product quantization. We show that strong correlation among the feature dimensions in the FV and the VLAD may not exist, which renders feature selection a natural choice. We also show that, many dimensions in FV/VLAD are noise. Throwing them away using feature selection is better than compressing them and useful dimensions altogether using feature compression methods. To choose features, we propose an efficient importance sorting algorithm considering both the supervised and unsupervised cases, for visual recognition and image retrieval, respectively. Combining with the 1-bit quantization, feature selection has achieved both higher accuracy and less computational cost than feature compression methods, such as product quantization, on the FV and the VLAD image representations.In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors (VLAD), have achieved state-of-the-art results. However, the combination of the large numbers of examples and high-dimensional vectors necessitates dimensionality reduction, in order to reduce its storage and CPU costs to a reasonable range. In spite of the popularity of various feature compression methods, this paper shows that the feature (dimension) selection is a better choice for high-dimensional FV/VLAD than the feature (dimension) compression methods, e.g., product quantization. We show that strong correlation among the feature dimensions in the FV and the VLAD may not exist, which renders feature selection a natural choice. We also show that, many dimensions in FV/VLAD are noise. Throwing them away using feature selection is better than compressing them and useful dimensions altogether using feature compression methods. To choose features, we propose an efficient importance sorting algorithm considering both the supervised and unsupervised cases, for visual recognition and image retrieval, respectively. Combining with the 1-bit quantization, feature selection has achieved both higher accuracy and less computational cost than feature compression methods, such as product quantization, on the FV and the VLAD image representations.
Author	Yu Zhang Jianxin Wu Jianfei Cai
Author_xml	– sequence: 1 givenname: Yu surname: Zhang fullname: Zhang, Yu – sequence: 2 givenname: Jianxin surname: Wu fullname: Wu, Jianxin – sequence: 3 givenname: Jianfei surname: Cai fullname: Cai, Jianfei
BackLink	https://www.ncbi.nlm.nih.gov/pubmed/27046897$$D View this record in MEDLINE/PubMed
BookMark	eNp9kUtv1DAUhS1URNuBPRISisSmmwzXj8TxEk0pHWkkEBS20R3nZvAoiQfbQeLf12WGLrpg5Ye-c659ziU7m_xEjL3msOQczPu79ZelAF4vRaWMrOEZu-BG8RJAibO8h0qXmitzzi5j3ANwVfH6BTsXGlTdGH3B9is_HtCm4isdAkWaEibnp8L3xa3b_Syv3UhTzDc4FDeEaQ5U_CCbfIhF70OxwbCj8pvFgYr1iDvKRtbvJvfXBacun1Nw9BuHl-x5j0OkV6d1wb7ffLxb3Zabz5_Wqw-b0kqlU2kF1l1ltkaISjbcGCAr-8Y0UoHaIjXUoYVO6kpJJa2wHTY9IOkaK05bkAt2dfQ9BP9rppja0UVLw4AT-Tm2vAEttFaNyei7J-jezyH_NVO6yckpnh-xYG9P1LwdqWsPwY0Y_rT_UswAHAEbfIyB-keEQ_tQVJuLah-Kak9FZUn9RGLdMfoU0A3_E745Ch0RPc7RSlVCCnkPrv6e9A
CODEN	IIPRE4
CitedBy_id	crossref_primary_10_3233_JIFS_189319 crossref_primary_10_1111_exsy_12459 crossref_primary_10_1186_s13640_017_0176_3 crossref_primary_10_1145_3314051 crossref_primary_10_1109_ACCESS_2019_2894366 crossref_primary_10_1016_j_patcog_2022_108541 crossref_primary_10_3390_informatics10020051 crossref_primary_10_3390_math13111819 crossref_primary_10_1016_j_patcog_2019_107167 crossref_primary_10_1109_TCSVT_2017_2671899 crossref_primary_10_1007_s00500_019_04355_y crossref_primary_10_1016_j_energy_2025_138466 crossref_primary_10_3390_info9020038 crossref_primary_10_1109_TIP_2023_3348992 crossref_primary_10_1109_TIP_2016_2605305
Cites_doi	10.1109/CVPR.2013.69 10.1109/CVPR.2012.6247943 10.1023/B:VISI.0000027790.02288.f2 10.1109/TKDE.2014.2320728 10.1109/CVPR.2013.388 10.1109/CVPR.2008.4587635 10.1109/CVPR.2007.383172 10.1017/CBO9780511809071 10.1109/CVPR.2006.68 10.1007/s11263-013-0636-x 10.1023/A:1025667309714 10.1109/TPAMI.2006.134 10.1109/TIP.2014.2344296 10.1109/CVPR.2013.207 10.1109/TPAMI.2010.57 10.1007/978-3-319-10584-0_26 10.1109/TIP.2016.2531289 10.1109/CVPR.2013.389 10.1109/CVPR.2010.5540009 10.1109/TPAMI.2005.159 10.1109/TIP.2014.2332396 10.1109/CVPR.2014.121 10.1109/CVPR.2012.6248090 10.1109/CVPR.2014.330 10.1109/TNNLS.2015.2458986 10.1109/TPAMI.2011.235 10.1109/CVPR.2010.5540039 10.1109/CVPR.2011.5995477 10.1109/CVPR.2011.5995504 10.1109/TPAMI.2015.2456899 10.1109/TIP.2008.2001050 10.1007/s11263-009-0285-2 10.1109/CVPR.2011.5995432 10.1109/TIP.2010.2093906 10.1109/CVPR.2013.379 10.1109/CVPR.2014.417 10.1109/JPROC.2012.2188013 10.1109/CVPR.2010.5539970
ContentType	Journal Article
Copyright	Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2016
Copyright_xml	– notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2016
DBID	97E RIA RIE AAYXX CITATION NPM 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8
DOI	10.1109/TIP.2016.2549360
DatabaseName	IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE 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 MEDLINE - Academic Technology Research Database
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 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	Applied Sciences Engineering
EISSN	1941-0042
EndPage	2419
ExternalDocumentID	4046914561 27046897 10_1109_TIP_2016_2549360 7445232
Genre	orig-research Research Support, Non-U.S. Gov't Journal Article
GrantInformation_xml	– fundername: Singapore Minister for Education through the Academic Research Funding Tier-1 grantid: RG138/14 – fundername: National Natural Science Foundation of China grantid: 61422203 funderid: 10.13039/501100001809
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~ O9- OCL P2P RIA RIE RNS TAE TN5 VH1 AAYXX CITATION NPM RIG 7SC 7SP 8FD JQ2 L7M L~C L~D 7X8
ID	FETCH-LOGICAL-c347t-c2a6d59b9225381990ec3f8983404bae8edac0d3754343c2cda8f0ae76a51eb03
IEDL.DBID	RIE
ISICitedReferencesCount	1842
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000374889800001&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 16:40:10 EDT 2025 Mon Jun 30 10:25:49 EDT 2025 Mon Jul 21 06:01:45 EDT 2025 Tue Nov 18 21:52:55 EST 2025 Sat Nov 29 03:21:02 EST 2025 Tue Aug 26 16:43:04 EDT 2025
IsPeerReviewed	true
IsScholarly	true
Issue	5
Keywords	large scale Feature selection image representation
Language	English
License	https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c347t-c2a6d59b9225381990ec3f8983404bae8edac0d3754343c2cda8f0ae76a51eb03
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23
PMID	27046897
PQID	1787144125
PQPubID	85429
PageCount	13
ParticipantIDs	proquest_miscellaneous_1807277489 proquest_journals_1787144125 pubmed_primary_27046897 crossref_citationtrail_10_1109_TIP_2016_2549360 crossref_primary_10_1109_TIP_2016_2549360 ieee_primary_7445232
PublicationCentury	2000
PublicationDate	2016-05-01
PublicationDateYYYYMMDD	2016-05-01
PublicationDate_xml	– month: 05 year: 2016 text: 2016-05-01 day: 01
PublicationDecade	2010
PublicationPlace	United States
PublicationPlace_xml	– name: United States – name: New York
PublicationTitle	IEEE transactions on image processing
PublicationTitleAbbrev	TIP
PublicationTitleAlternate	IEEE Trans Image Process
PublicationYear	2016
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	ref13 ref12 ref15 nie (ref54) 2010 ref14 ref52 ref55 ref11 ref10 ref17 ref16 qian (ref31) 2013 ref19 deng (ref1) 2009 monteiro (ref18) 2014 berg (ref46) 2015 ref51 ref50 ref48 ref47 ref41 ref43 ref49 ref8 chatfield (ref42) 2010 krizhevsky (ref7) 2012 ref9 ref4 ref3 ref6 ref5 ref40 weiss (ref21) 2008 tan (ref22) 2010 fan (ref44) 2008; 9 ref35 ref34 ref37 ref36 ref2 ref39 ref38 li (ref30) 2012 ref23 ref26 ref25 ref20 he (ref29) 2005 fleuret (ref24) 2004; 5 yang (ref33) 2011 gao (ref53) 2011 ref28 ref27 obozinski (ref32) 2006 everingham (ref45) 2015
References_xml	– start-page: 248 year: 2009 ident: ref1 article-title: ImageNet: A large-scale hierarchical image database publication-title: Proc IEEE Int Conf Comput Vis Pattern Recognit – volume: 9 start-page: 1871 year: 2008 ident: ref44 article-title: LIBLINEAR: A library for large linear classification publication-title: J Mach Learn Res – ident: ref15 doi: 10.1109/CVPR.2013.69 – ident: ref11 doi: 10.1109/CVPR.2012.6247943 – ident: ref48 doi: 10.1023/B:VISI.0000027790.02288.f2 – ident: ref34 doi: 10.1109/TKDE.2014.2320728 – ident: ref12 doi: 10.1109/CVPR.2013.388 – ident: ref51 doi: 10.1109/CVPR.2008.4587635 – start-page: 1813 year: 2010 ident: ref54 article-title: Efficient and robust feature selection via joint $\ell _{2,1}$ -norms minimization publication-title: Proc Adv Neural Inf Process Syst – ident: ref50 doi: 10.1109/CVPR.2007.383172 – ident: ref36 doi: 10.1017/CBO9780511809071 – ident: ref35 doi: 10.1109/CVPR.2006.68 – ident: ref2 doi: 10.1007/s11263-013-0636-x – start-page: 1097 year: 2012 ident: ref7 article-title: ImageNet classification with deep convolutional neural networks publication-title: Proc Adv Neural Inf Process Syst – start-page: 1621 year: 2013 ident: ref31 article-title: Robust unsupervised feature selection publication-title: Proc 23rd Int Joint Conf Artif Intell – ident: ref28 doi: 10.1023/A:1025667309714 – ident: ref6 doi: 10.1109/TPAMI.2006.134 – ident: ref43 doi: 10.1109/TIP.2014.2344296 – volume: 5 start-page: 1531 year: 2004 ident: ref24 article-title: Fast binary feature selection with conditional mutual information publication-title: J Mach Learn Res – ident: ref4 doi: 10.1109/CVPR.2013.207 – ident: ref10 doi: 10.1109/TPAMI.2010.57 – start-page: 1753 year: 2008 ident: ref21 article-title: Spectral hashing publication-title: Proc Adv Neural Inf Process Syst – ident: ref8 doi: 10.1007/978-3-319-10584-0_26 – ident: ref39 doi: 10.1109/TIP.2016.2531289 – ident: ref16 doi: 10.1109/CVPR.2013.389 – ident: ref37 doi: 10.1109/CVPR.2010.5540009 – ident: ref23 doi: 10.1109/TPAMI.2005.159 – ident: ref27 doi: 10.1109/TIP.2014.2332396 – ident: ref19 doi: 10.1109/CVPR.2014.121 – ident: ref40 doi: 10.1109/CVPR.2012.6248090 – ident: ref38 doi: 10.1109/CVPR.2014.330 – start-page: 1 year: 2006 ident: ref32 article-title: Multi-task feature selection publication-title: Proc Adv Neural Inf Process Syst – start-page: 1026 year: 2012 ident: ref30 article-title: Unsupervised feature selection using nonnegative spectral analysis publication-title: Proc 26th AAAI Conf Artif Intell – ident: ref17 doi: 10.1109/TNNLS.2015.2458986 – year: 2015 ident: ref45 publication-title: The PASCAL Visual Object Classes Challenge 2007 – start-page: 1047 year: 2010 ident: ref22 article-title: Learning sparse SVM for feature selection on very high dimensional datasets publication-title: Proc Int Conf Mach Learn – ident: ref3 doi: 10.1109/TPAMI.2011.235 – ident: ref20 doi: 10.1109/CVPR.2010.5540039 – ident: ref52 doi: 10.1109/CVPR.2011.5995477 – ident: ref9 doi: 10.1109/CVPR.2011.5995504 – start-page: 76.1 year: 2010 ident: ref42 article-title: The devil is in the details: An evaluation of recent feature encoding methods publication-title: Proc Brit Mach Vis Conf – ident: ref5 doi: 10.1109/TPAMI.2015.2456899 – ident: ref25 doi: 10.1109/TIP.2008.2001050 – year: 2015 ident: ref46 publication-title: ImageNet Large Scale Visual Recognition Challenge 2010 (ILSVRC2010) – ident: ref49 doi: 10.1007/s11263-009-0285-2 – ident: ref14 doi: 10.1109/CVPR.2011.5995432 – ident: ref26 doi: 10.1109/TIP.2010.2093906 – start-page: 2072 year: 2011 ident: ref53 article-title: Discriminative learning of relaxed hierarchy for large-scale visual recognition publication-title: Proc IEEE Int Conf Comput Vis – ident: ref13 doi: 10.1109/CVPR.2013.379 – start-page: 507 year: 2005 ident: ref29 article-title: Laplacian score for feature selection publication-title: Proc Adv Neural Inf Process Syst – ident: ref55 doi: 10.1109/CVPR.2014.417 – start-page: 1589 year: 2011 ident: ref33 article-title: $\ell _{2,1}$ -norm regularized discriminative feature selection for unsupervised learning publication-title: Proc 22nd Int Joint Conf Artif Intell – ident: ref41 doi: 10.1109/JPROC.2012.2188013 – start-page: 541 year: 2014 ident: ref18 article-title: Coding mode decision algorithm for binary descriptor coding publication-title: Proc 22nd Eur Signal Process Conf (EUSIPCO) – ident: ref47 doi: 10.1109/CVPR.2010.5539970
SSID	ssj0014516
Score	2.6879506
Snippet	In large-scale visual recognition and image retrieval tasks, feature vectors, such as Fisher vector (FV) or the vector of locally aggregated descriptors...
SourceID	proquest pubmed crossref ieee
SourceType	Aggregation Database Index Database Enrichment Source Publisher
StartPage	2407
SubjectTerms	Correlation Digital imaging feature selection Image coding Image recognition image representation Image retrieval large scale Quantization (signal) Sorting Visualization
Title	Compact Representation of High-Dimensional Feature Vectors for Large-Scale Image Recognition and Retrieval
URI	https://ieeexplore.ieee.org/document/7445232 https://www.ncbi.nlm.nih.gov/pubmed/27046897 https://www.proquest.com/docview/1787144125 https://www.proquest.com/docview/1807277489
Volume	25
WOSCitedRecordID	wos000374889800001&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/eLvHCXMwlV3da9UwFD_M4YM-ON10VueI4ItgdnObtEkehzocyBhzyn0raZKCMnvlfvj3e06alj04wbeWpm3gfOScnJPfD-BNKVtJMGqcsFmIwqzkrZSWV2quQyl8V7cD2YS-uDCLhb3cgXfTWZgYY2o-iyd0mWr5Yem3tFU200ph3oQO957W9XBWa6oYEOFsqmxWmmsM-8eSpLCz6_NL6uGqTygZkjWRv5Ua80JDSE-3VqNEr3J3pJlWnLO9_5vrY3iUI0t2OqjCE9iJ_T7s5SiTZRte78PDWxCEB_AjOQS_YVepJTafROrZsmPUAsI_EPr_gNzBKFzcriL7lnb61wzjXfaZOsn5F5R0ZOc_0Tmxq7ElCb_i-oD3xNmFCv0Uvp59vH7_iWf-Be6l0hvuS1eHyrYWbZ4SOyuil52xRiqhWhdNDM6LQCS6Uklf-uBMJ1zUtavmsRXyGez2yz4-B-a0Qcv3Qc6rqJRWbdlhLoqpeVCi7WxdwGyUQ-MzODlxZNw0KUkRtkEhNiTEJguxgLfTG78GYI5_jD0gAU3jsmwKOBpF3WTLXTdz9GCUZJZVAa-nx2hzVEhxfVxucYwRGPYRbk8Bh4OKTN8eNevF3__5Eh7QzIaWySPY3ay28RXc978339erY1TshTlOiv0H9jLwOg
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lb9QwEB5VBQk4UGgpBAoYiQsS7nptJ06OCKi6YllVZUG9RY7tSCDIon3w-5lxnKgHQOKWKE5iaR6e8Yy_D-ClVI0iGDVO2CxEYSZ5o1TFcz01XgrXFk1PNmEWi_LqqrrYg9fjWZgQQmw-C6d0GWv5fuV2tFU2MVpj3oQO90autRT9aa2xZkCUs7G2mRtuMPAfipKimixnF9TFVZxSOqQKon-TBjPDkrCerq1HkWDl77FmXHPODv5vtvfgboot2ZteGe7DXugO4SDFmSxZ8eYQ7lwDITyCb9EluC27jE2x6SxSx1YtoyYQ_o7w_3vsDkYB424d2Je4179hGPGyOfWS808o68BmP9A9scuhKQm_YjuP98TahSr9AD6fvV--PeeJgYE7pc2WO2kLn1dNhVZPqV0lglNtWZVKC93YUAZvnfBEo6u0ctJ5W7bCBlPYfBoaoY5hv1t14REwa0q0fefVNA9aG93IFrNRTM69Fk1bFRlMBjnULsGTE0vG9zqmKaKqUYg1CbFOQszg1fjGzx6a4x9jj0hA47gkmwxOBlHXyXY39RR9GKWZMs_gxfgYrY5KKbYLqx2OKQUGfoTck8HDXkXGbw-a9fjP_3wOt86XH-f1fLb48ARu0yz7BsoT2N-ud-Ep3HS_tl8362dRvX8D70nymQ
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=Compact+Representation+of+High-Dimensional+Feature+Vectors+for+Large-Scale+Image+Recognition+and+Retrieval&rft.jtitle=IEEE+transactions+on+image+processing&rft.au=Yu+Zhang&rft.au=Jianxin+Wu&rft.au=Jianfei+Cai&rft.date=2016-05-01&rft.pub=IEEE&rft.issn=1057-7149&rft.volume=25&rft.issue=5&rft.spage=2407&rft.epage=2419&rft_id=info:doi/10.1109%2FTIP.2016.2549360&rft_id=info%3Apmid%2F27046897&rft.externalDocID=7445232
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