Multiple random walk and its application in content-based image retrieval

Jingrui He, Hanghang Tong, Mingjing Li, Ma Wei-Ying, Changshui Zhang

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

7 Citations (Scopus)

Abstract

In this paper, we propose a transductive learning method for content-based image retrieval: Multiple Random Walk (MRW). Its basic idea is to construct two generative models by means of Markov random walks, one for images relevant to the query concept and the other for the irrelevant ones. The goal is to obtain the likelihood functions of both classes. Firstly, MRW generates two random walks with virtual absorbing boundaries, and uses the absorbing probabilities as the initial estimation of the likelihood functions. Then it refines the two random walks through an EMlike iterative procedure in order to get more accurate estimation of the likelihood functions. Class priors are also obtained in this procedure. Finally, MRW ranks all the unlabeled images in the database according to their posterior probabilities of being relevant. By using both labeled and unlabeled data, MRW can be seen as a transductive learning method, which has been demonstrated to outperform inductive ones by previous research work. Systematic experiments on a general-purpose image database consisting of 5,000 Corel images demonstrate the superiority of MRW over state-of-the-art techniques.

Original languageEnglish (US)
Title of host publicationMIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005
PublisherAssociation for Computing Machinery, Inc
Pages151-158
Number of pages8
ISBN (Print)1595932445, 9781595932440
StatePublished - Nov 10 2005
Externally publishedYes
Event7th ACM SIGMM International Workshop on Multimedia Information Retrieval, MIR 2005 - Singapore, Singapore
Duration: Nov 10 2005Nov 11 2005

Other

Other7th ACM SIGMM International Workshop on Multimedia Information Retrieval, MIR 2005
CountrySingapore
CitySingapore
Period11/10/0511/11/05

Fingerprint

Image retrieval
Experiments

Keywords

  • Class prior
  • Generative model
  • Image retrieval
  • Likelihood function
  • Markov random walk
  • Relevance feedback.

ASJC Scopus subject areas

  • Computer Science Applications
  • Information Systems
  • Signal Processing
  • Software
  • Media Technology

Cite this

He, J., Tong, H., Li, M., Wei-Ying, M., & Zhang, C. (2005). Multiple random walk and its application in content-based image retrieval. In MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005 (pp. 151-158). Association for Computing Machinery, Inc.

Multiple random walk and its application in content-based image retrieval. / He, Jingrui; Tong, Hanghang; Li, Mingjing; Wei-Ying, Ma; Zhang, Changshui.

MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005. Association for Computing Machinery, Inc, 2005. p. 151-158.

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

He, J, Tong, H, Li, M, Wei-Ying, M & Zhang, C 2005, Multiple random walk and its application in content-based image retrieval. in MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005. Association for Computing Machinery, Inc, pp. 151-158, 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, MIR 2005, Singapore, Singapore, 11/10/05.
He J, Tong H, Li M, Wei-Ying M, Zhang C. Multiple random walk and its application in content-based image retrieval. In MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005. Association for Computing Machinery, Inc. 2005. p. 151-158
He, Jingrui ; Tong, Hanghang ; Li, Mingjing ; Wei-Ying, Ma ; Zhang, Changshui. / Multiple random walk and its application in content-based image retrieval. MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005. Association for Computing Machinery, Inc, 2005. pp. 151-158
@inproceedings{95301425a4664c8c93330218adad1259,
title = "Multiple random walk and its application in content-based image retrieval",
abstract = "In this paper, we propose a transductive learning method for content-based image retrieval: Multiple Random Walk (MRW). Its basic idea is to construct two generative models by means of Markov random walks, one for images relevant to the query concept and the other for the irrelevant ones. The goal is to obtain the likelihood functions of both classes. Firstly, MRW generates two random walks with virtual absorbing boundaries, and uses the absorbing probabilities as the initial estimation of the likelihood functions. Then it refines the two random walks through an EMlike iterative procedure in order to get more accurate estimation of the likelihood functions. Class priors are also obtained in this procedure. Finally, MRW ranks all the unlabeled images in the database according to their posterior probabilities of being relevant. By using both labeled and unlabeled data, MRW can be seen as a transductive learning method, which has been demonstrated to outperform inductive ones by previous research work. Systematic experiments on a general-purpose image database consisting of 5,000 Corel images demonstrate the superiority of MRW over state-of-the-art techniques.",
keywords = "Class prior, Generative model, Image retrieval, Likelihood function, Markov random walk, Relevance feedback.",
author = "Jingrui He and Hanghang Tong and Mingjing Li and Ma Wei-Ying and Changshui Zhang",
year = "2005",
month = "11",
day = "10",
language = "English (US)",
isbn = "1595932445",
pages = "151--158",
booktitle = "MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005",
publisher = "Association for Computing Machinery, Inc",

}

TY - GEN

T1 - Multiple random walk and its application in content-based image retrieval

AU - He, Jingrui

AU - Tong, Hanghang

AU - Li, Mingjing

AU - Wei-Ying, Ma

AU - Zhang, Changshui

PY - 2005/11/10

Y1 - 2005/11/10

N2 - In this paper, we propose a transductive learning method for content-based image retrieval: Multiple Random Walk (MRW). Its basic idea is to construct two generative models by means of Markov random walks, one for images relevant to the query concept and the other for the irrelevant ones. The goal is to obtain the likelihood functions of both classes. Firstly, MRW generates two random walks with virtual absorbing boundaries, and uses the absorbing probabilities as the initial estimation of the likelihood functions. Then it refines the two random walks through an EMlike iterative procedure in order to get more accurate estimation of the likelihood functions. Class priors are also obtained in this procedure. Finally, MRW ranks all the unlabeled images in the database according to their posterior probabilities of being relevant. By using both labeled and unlabeled data, MRW can be seen as a transductive learning method, which has been demonstrated to outperform inductive ones by previous research work. Systematic experiments on a general-purpose image database consisting of 5,000 Corel images demonstrate the superiority of MRW over state-of-the-art techniques.

AB - In this paper, we propose a transductive learning method for content-based image retrieval: Multiple Random Walk (MRW). Its basic idea is to construct two generative models by means of Markov random walks, one for images relevant to the query concept and the other for the irrelevant ones. The goal is to obtain the likelihood functions of both classes. Firstly, MRW generates two random walks with virtual absorbing boundaries, and uses the absorbing probabilities as the initial estimation of the likelihood functions. Then it refines the two random walks through an EMlike iterative procedure in order to get more accurate estimation of the likelihood functions. Class priors are also obtained in this procedure. Finally, MRW ranks all the unlabeled images in the database according to their posterior probabilities of being relevant. By using both labeled and unlabeled data, MRW can be seen as a transductive learning method, which has been demonstrated to outperform inductive ones by previous research work. Systematic experiments on a general-purpose image database consisting of 5,000 Corel images demonstrate the superiority of MRW over state-of-the-art techniques.

KW - Class prior

KW - Generative model

KW - Image retrieval

KW - Likelihood function

KW - Markov random walk

KW - Relevance feedback.

UR - http://www.scopus.com/inward/record.url?scp=34547398988&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34547398988&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:34547398988

SN - 1595932445

SN - 9781595932440

SP - 151

EP - 158

BT - MIR 2005 - Proceedings of the 7th ACM SIGMM International Workshop on Multimedia Information Retrieval, Co-located with ACM Multimedia 2005

PB - Association for Computing Machinery, Inc

ER -