Independent feature analysis for image retrieval

Jing Peng, Bir Bhanu

Research output: Contribution to journalArticleResearchpeer-review

3 Citations (Scopus)

Abstract

Content-based image retrieval methods based on the Euclidean metric expect the feature space to be isotropic. They suffer from unequal differential relevance of features in computing the similarity between images in the input feature space. We propose a learning method that attempts to overcome this limitation by capturing local differential relevance of features based on user feedback. This feedback, in the form of accept or reject examples generated in respones to a query image, is used to locally estimate the strength of features along each dimension while taking into considertaion the correlation between features. This results in local neighborhood that are constricted along feature dimensions and that are most relevant, while elongated along less relevant ones. In addition to exploring and exploiting local principal information, the system seeks a global for efficient independent feature analysis by combining such local information. We provide experimental results that demonstrate the efficacy of our technique using both simulated and real-world data.

Original languageEnglish
Pages (from-to)63-74
Number of pages12
JournalPattern Recognition Letters
Volume22
Issue number1
DOIs
StatePublished - 28 Feb 2001

Fingerprint

Image retrieval
Feedback

Keywords

  • Feature analysis
  • Image retrieval

Cite this

Peng, Jing ; Bhanu, Bir. / Independent feature analysis for image retrieval. In: Pattern Recognition Letters. 2001 ; Vol. 22, No. 1. pp. 63-74.
@article{32ef8e5b2e5f407dbdbc1a896cb66b2e,
title = "Independent feature analysis for image retrieval",
abstract = "Content-based image retrieval methods based on the Euclidean metric expect the feature space to be isotropic. They suffer from unequal differential relevance of features in computing the similarity between images in the input feature space. We propose a learning method that attempts to overcome this limitation by capturing local differential relevance of features based on user feedback. This feedback, in the form of accept or reject examples generated in respones to a query image, is used to locally estimate the strength of features along each dimension while taking into considertaion the correlation between features. This results in local neighborhood that are constricted along feature dimensions and that are most relevant, while elongated along less relevant ones. In addition to exploring and exploiting local principal information, the system seeks a global for efficient independent feature analysis by combining such local information. We provide experimental results that demonstrate the efficacy of our technique using both simulated and real-world data.",
keywords = "Feature analysis, Image retrieval",
author = "Jing Peng and Bir Bhanu",
year = "2001",
month = "2",
day = "28",
doi = "10.1016/S0167-8655(00)00100-8",
language = "English",
volume = "22",
pages = "63--74",
journal = "Pattern Recognition Letters",
issn = "0167-8655",
publisher = "Elsevier",
number = "1",

}

Independent feature analysis for image retrieval. / Peng, Jing; Bhanu, Bir.

In: Pattern Recognition Letters, Vol. 22, No. 1, 28.02.2001, p. 63-74.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Independent feature analysis for image retrieval

AU - Peng, Jing

AU - Bhanu, Bir

PY - 2001/2/28

Y1 - 2001/2/28

N2 - Content-based image retrieval methods based on the Euclidean metric expect the feature space to be isotropic. They suffer from unequal differential relevance of features in computing the similarity between images in the input feature space. We propose a learning method that attempts to overcome this limitation by capturing local differential relevance of features based on user feedback. This feedback, in the form of accept or reject examples generated in respones to a query image, is used to locally estimate the strength of features along each dimension while taking into considertaion the correlation between features. This results in local neighborhood that are constricted along feature dimensions and that are most relevant, while elongated along less relevant ones. In addition to exploring and exploiting local principal information, the system seeks a global for efficient independent feature analysis by combining such local information. We provide experimental results that demonstrate the efficacy of our technique using both simulated and real-world data.

AB - Content-based image retrieval methods based on the Euclidean metric expect the feature space to be isotropic. They suffer from unequal differential relevance of features in computing the similarity between images in the input feature space. We propose a learning method that attempts to overcome this limitation by capturing local differential relevance of features based on user feedback. This feedback, in the form of accept or reject examples generated in respones to a query image, is used to locally estimate the strength of features along each dimension while taking into considertaion the correlation between features. This results in local neighborhood that are constricted along feature dimensions and that are most relevant, while elongated along less relevant ones. In addition to exploring and exploiting local principal information, the system seeks a global for efficient independent feature analysis by combining such local information. We provide experimental results that demonstrate the efficacy of our technique using both simulated and real-world data.

KW - Feature analysis

KW - Image retrieval

UR - http://www.scopus.com/inward/record.url?scp=0034746274&partnerID=8YFLogxK

U2 - 10.1016/S0167-8655(00)00100-8

DO - 10.1016/S0167-8655(00)00100-8

M3 - Article

VL - 22

SP - 63

EP - 74

JO - Pattern Recognition Letters

JF - Pattern Recognition Letters

SN - 0167-8655

IS - 1

ER -