Probabilistic feature relevance learning for content-based image retrieval

Jing Peng, Bir Bhanu, Shan Qing

Research output: Contribution to journalArticleResearchpeer-review

119 Citations (Scopus)

Abstract

Most of the current image retrieval systems use `one-shot' queries to a database to retrieve similar images. Typically a K-nearest neighbor kind of algorithm is used, where weights measuring feature importance along each input dimension remain fixed (or manually tweaked by the user), in the computation of a given similarity metric. However, the similarity does not vary with equal strength or in the same proportion in all directions in the feature space emanating from the query image. The manual adjustment of these weights is time consuming and exhausting. Moreover, it requires a very sophisticated user. In this paper, we present a novel probabilistic method that enables image retrieval procedures to automatically capture feature relevance based on user's feedback and that is highly adaptive to query locations. Experimental results are presented that demonstrate the efficacy of our technique using both simulated and real-world data.

Original languageEnglish
Pages (from-to)150-164
Number of pages15
JournalComputer Vision and Image Understanding
Volume75
Issue number1
DOIs
StatePublished - 1 Jan 1999

Fingerprint

Image retrieval
Feedback

Cite this

@article{94d19f3943204170a14ce046fb11bd7c,
title = "Probabilistic feature relevance learning for content-based image retrieval",
abstract = "Most of the current image retrieval systems use `one-shot' queries to a database to retrieve similar images. Typically a K-nearest neighbor kind of algorithm is used, where weights measuring feature importance along each input dimension remain fixed (or manually tweaked by the user), in the computation of a given similarity metric. However, the similarity does not vary with equal strength or in the same proportion in all directions in the feature space emanating from the query image. The manual adjustment of these weights is time consuming and exhausting. Moreover, it requires a very sophisticated user. In this paper, we present a novel probabilistic method that enables image retrieval procedures to automatically capture feature relevance based on user's feedback and that is highly adaptive to query locations. Experimental results are presented that demonstrate the efficacy of our technique using both simulated and real-world data.",
author = "Jing Peng and Bir Bhanu and Shan Qing",
year = "1999",
month = "1",
day = "1",
doi = "10.1006/cviu.1999.0770",
language = "English",
volume = "75",
pages = "150--164",
journal = "Computer Vision and Image Understanding",
issn = "1077-3142",
publisher = "Academic Press Inc.",
number = "1",

}

Probabilistic feature relevance learning for content-based image retrieval. / Peng, Jing; Bhanu, Bir; Qing, Shan.

In: Computer Vision and Image Understanding, Vol. 75, No. 1, 01.01.1999, p. 150-164.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Probabilistic feature relevance learning for content-based image retrieval

AU - Peng, Jing

AU - Bhanu, Bir

AU - Qing, Shan

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Most of the current image retrieval systems use `one-shot' queries to a database to retrieve similar images. Typically a K-nearest neighbor kind of algorithm is used, where weights measuring feature importance along each input dimension remain fixed (or manually tweaked by the user), in the computation of a given similarity metric. However, the similarity does not vary with equal strength or in the same proportion in all directions in the feature space emanating from the query image. The manual adjustment of these weights is time consuming and exhausting. Moreover, it requires a very sophisticated user. In this paper, we present a novel probabilistic method that enables image retrieval procedures to automatically capture feature relevance based on user's feedback and that is highly adaptive to query locations. Experimental results are presented that demonstrate the efficacy of our technique using both simulated and real-world data.

AB - Most of the current image retrieval systems use `one-shot' queries to a database to retrieve similar images. Typically a K-nearest neighbor kind of algorithm is used, where weights measuring feature importance along each input dimension remain fixed (or manually tweaked by the user), in the computation of a given similarity metric. However, the similarity does not vary with equal strength or in the same proportion in all directions in the feature space emanating from the query image. The manual adjustment of these weights is time consuming and exhausting. Moreover, it requires a very sophisticated user. In this paper, we present a novel probabilistic method that enables image retrieval procedures to automatically capture feature relevance based on user's feedback and that is highly adaptive to query locations. Experimental results are presented that demonstrate the efficacy of our technique using both simulated and real-world data.

UR - http://www.scopus.com/inward/record.url?scp=0032636367&partnerID=8YFLogxK

U2 - 10.1006/cviu.1999.0770

DO - 10.1006/cviu.1999.0770

M3 - Article

VL - 75

SP - 150

EP - 164

JO - Computer Vision and Image Understanding

JF - Computer Vision and Image Understanding

SN - 1077-3142

IS - 1

ER -