Discriminative learning of local image descriptors

Matthew A Brown, Gang Hua, S Winder

Research output: Contribution to journalArticle

  • 179 Citations

Abstract

In this paper, we explore methods for learning local image descriptors from training data. We describe a set of building blocks for constructing descriptors which can be combined together and jointly optimized so as to minimize the error of a nearest-neighbor classifier. We consider both linear and nonlinear transforms with dimensionality reduction, and make use of discriminant learning techniques such as Linear Discriminant Analysis (LDA) and Powell minimization to solve for the parameters. Using these techniques, we obtain descriptors that exceed state-of-the-art performance with low dimensionality. In addition to new experiments and recommendations for descriptor learning, we are also making available a new and realistic ground truth data set based on multiview stereo data.
LanguageEnglish
Pages43-57
Number of pages15
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume33
Issue number1
Early online date25 Feb 2010
DOIs
StatusPublished - Jan 2011

Fingerprint

Discriminant analysis
Descriptors
Classifiers
Experiments
Dimensionality Reduction
Discriminant Analysis
Discriminant
Building Blocks
Dimensionality
Recommendations
Nearest Neighbor
Exceed
Classifier
Transform
Minimise
Learning
Experiment

Cite this

Brown, M. A., Hua, G., & Winder, S. (2011). Discriminative learning of local image descriptors. DOI: 10.1109/TPAMI.2010.54

Discriminative learning of local image descriptors. / Brown, Matthew A; Hua, Gang; Winder, S.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 33, No. 1, 01.2011, p. 43-57.

Research output: Contribution to journalArticle

Brown, Matthew A ; Hua, Gang ; Winder, S. / Discriminative learning of local image descriptors. In: IEEE Transactions on Pattern Analysis and Machine Intelligence. 2011 ; Vol. 33, No. 1. pp. 43-57
@article{9d217289d1c74d0d9085ab1bf98f006e,
title = "Discriminative learning of local image descriptors",
abstract = "In this paper, we explore methods for learning local image descriptors from training data. We describe a set of building blocks for constructing descriptors which can be combined together and jointly optimized so as to minimize the error of a nearest-neighbor classifier. We consider both linear and nonlinear transforms with dimensionality reduction, and make use of discriminant learning techniques such as Linear Discriminant Analysis (LDA) and Powell minimization to solve for the parameters. Using these techniques, we obtain descriptors that exceed state-of-the-art performance with low dimensionality. In addition to new experiments and recommendations for descriptor learning, we are also making available a new and realistic ground truth data set based on multiview stereo data.",
author = "Brown, {Matthew A} and Gang Hua and S Winder",
year = "2011",
month = "1",
doi = "10.1109/TPAMI.2010.54",
language = "English",
volume = "33",
pages = "43--57",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE",
number = "1",

}

TY - JOUR

T1 - Discriminative learning of local image descriptors

AU - Brown,Matthew A

AU - Hua,Gang

AU - Winder,S

PY - 2011/1

Y1 - 2011/1

N2 - In this paper, we explore methods for learning local image descriptors from training data. We describe a set of building blocks for constructing descriptors which can be combined together and jointly optimized so as to minimize the error of a nearest-neighbor classifier. We consider both linear and nonlinear transforms with dimensionality reduction, and make use of discriminant learning techniques such as Linear Discriminant Analysis (LDA) and Powell minimization to solve for the parameters. Using these techniques, we obtain descriptors that exceed state-of-the-art performance with low dimensionality. In addition to new experiments and recommendations for descriptor learning, we are also making available a new and realistic ground truth data set based on multiview stereo data.

AB - In this paper, we explore methods for learning local image descriptors from training data. We describe a set of building blocks for constructing descriptors which can be combined together and jointly optimized so as to minimize the error of a nearest-neighbor classifier. We consider both linear and nonlinear transforms with dimensionality reduction, and make use of discriminant learning techniques such as Linear Discriminant Analysis (LDA) and Powell minimization to solve for the parameters. Using these techniques, we obtain descriptors that exceed state-of-the-art performance with low dimensionality. In addition to new experiments and recommendations for descriptor learning, we are also making available a new and realistic ground truth data set based on multiview stereo data.

UR - http://www.scopus.com/inward/record.url?scp=78649324041&partnerID=8YFLogxK

UR - http://dx.doi.org/10.1109/TPAMI.2010.54

U2 - 10.1109/TPAMI.2010.54

DO - 10.1109/TPAMI.2010.54

M3 - Article

VL - 33

SP - 43

EP - 57

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

T2 - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 1

ER -