Adaptive multi-class metric content-based image retrieval

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

Relevance feedback methods for content-based image retrieval have shown promise in a variety of image database applications. These techniques assume two (relevant and irrelevant) class relevance feedback. While simple computationally, two class relevance feedback often becomes inadequate in providing suÆcient information to help rapidly improve retrieval performance. We propose a locally adaptive technique for content-based image retrieval that enables relevance feedback to take on multi-class form. We estimate a exible multi-class metric for computing retrievals based on Chi-squared distance analysis. As a result, local data distributions can be suÆciently exploited, whereby rapid performance improvement can be achieved. The eÆcacy of our method is validated and compared against other competing techniques using a number of real world data sets.

Original languageEnglish
Title of host publicationAdvances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings
EditorsRobert Laurini
PublisherSpringer Verlag
Pages407-418
Number of pages12
ISBN (Print)3540411771
DOIs
StatePublished - 1 Jan 2000
Event4th International Conference on Visual Information Systems, VISUAL 2000 - Lyon, France
Duration: 2 Nov 20004 Nov 2000

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1929
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other4th International Conference on Visual Information Systems, VISUAL 2000
CountryFrance
CityLyon
Period2/11/004/11/00

Fingerprint

Relevance Feedback
Content-based Image Retrieval
Image retrieval
Multi-class
Feedback
Metric
Retrieval
Adaptive Techniques
Chi-squared
Image Database
Data Distribution
Computing
Estimate
Class

Cite this

Peng, J. (2000). Adaptive multi-class metric content-based image retrieval. In R. Laurini (Ed.), Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings (pp. 407-418). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1929). Springer Verlag. https://doi.org/10.1007/3-540-40053-2_36
Peng, Jing. / Adaptive multi-class metric content-based image retrieval. Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings. editor / Robert Laurini. Springer Verlag, 2000. pp. 407-418 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{a164fe9f55ea40b69b7796f490a26ebb,
title = "Adaptive multi-class metric content-based image retrieval",
abstract = "Relevance feedback methods for content-based image retrieval have shown promise in a variety of image database applications. These techniques assume two (relevant and irrelevant) class relevance feedback. While simple computationally, two class relevance feedback often becomes inadequate in providing su{\AE}cient information to help rapidly improve retrieval performance. We propose a locally adaptive technique for content-based image retrieval that enables relevance feedback to take on multi-class form. We estimate a exible multi-class metric for computing retrievals based on Chi-squared distance analysis. As a result, local data distributions can be su{\AE}ciently exploited, whereby rapid performance improvement can be achieved. The e{\AE}cacy of our method is validated and compared against other competing techniques using a number of real world data sets.",
author = "Jing Peng",
year = "2000",
month = "1",
day = "1",
doi = "10.1007/3-540-40053-2_36",
language = "English",
isbn = "3540411771",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "407--418",
editor = "Robert Laurini",
booktitle = "Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings",

}

Peng, J 2000, Adaptive multi-class metric content-based image retrieval. in R Laurini (ed.), Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1929, Springer Verlag, pp. 407-418, 4th International Conference on Visual Information Systems, VISUAL 2000, Lyon, France, 2/11/00. https://doi.org/10.1007/3-540-40053-2_36

Adaptive multi-class metric content-based image retrieval. / Peng, Jing.

Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings. ed. / Robert Laurini. Springer Verlag, 2000. p. 407-418 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1929).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - Adaptive multi-class metric content-based image retrieval

AU - Peng, Jing

PY - 2000/1/1

Y1 - 2000/1/1

N2 - Relevance feedback methods for content-based image retrieval have shown promise in a variety of image database applications. These techniques assume two (relevant and irrelevant) class relevance feedback. While simple computationally, two class relevance feedback often becomes inadequate in providing suÆcient information to help rapidly improve retrieval performance. We propose a locally adaptive technique for content-based image retrieval that enables relevance feedback to take on multi-class form. We estimate a exible multi-class metric for computing retrievals based on Chi-squared distance analysis. As a result, local data distributions can be suÆciently exploited, whereby rapid performance improvement can be achieved. The eÆcacy of our method is validated and compared against other competing techniques using a number of real world data sets.

AB - Relevance feedback methods for content-based image retrieval have shown promise in a variety of image database applications. These techniques assume two (relevant and irrelevant) class relevance feedback. While simple computationally, two class relevance feedback often becomes inadequate in providing suÆcient information to help rapidly improve retrieval performance. We propose a locally adaptive technique for content-based image retrieval that enables relevance feedback to take on multi-class form. We estimate a exible multi-class metric for computing retrievals based on Chi-squared distance analysis. As a result, local data distributions can be suÆciently exploited, whereby rapid performance improvement can be achieved. The eÆcacy of our method is validated and compared against other competing techniques using a number of real world data sets.

UR - http://www.scopus.com/inward/record.url?scp=84959050043&partnerID=8YFLogxK

U2 - 10.1007/3-540-40053-2_36

DO - 10.1007/3-540-40053-2_36

M3 - Conference contribution

SN - 3540411771

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

SP - 407

EP - 418

BT - Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings

A2 - Laurini, Robert

PB - Springer Verlag

ER -

Peng J. Adaptive multi-class metric content-based image retrieval. In Laurini R, editor, Advances in Visual Information Systems - 4th International Conference, VISUAL 2000, Proceedings. Springer Verlag. 2000. p. 407-418. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/3-540-40053-2_36