Learning Local Image Descriptors

Simon A J Winder; Matthew Brown

doi:10.1109/CVPR.2007.382971

Learning Local Image Descriptors

Simon A J Winder, Matthew Brown

Department of Computer Science

Research output: Contribution to conference › Paper › peer-review

283 Citations (SciVal)

Abstract

In this paper we study interest point descriptors for image matching and 3D reconstruction. We examine the building blocks of descriptor algorithms and evaluate numerous combinations of components. Various published descriptors such as SIFT, GLOH, and Spin images can be cast into our framework. For each candidate algorithm we learn good choices for parameters using a training set consisting of patches from a multi-image 3D reconstruction where accurate ground-truth matches are known. The best descriptors were those with log polar histogramming regions and feature vectors constructed from rectified outputs of steerable quadrature filters. At a 95% detection rate these gave one third of the incorrect matches produced by SIFT.

Original language	English
DOIs	https://doi.org/10.1109/CVPR.2007.382971
Publication status	Published - Jun 2007
Event	CVPR '07: IEEE Conference on Computer Vision and Pattern Recognition, 2007 - Minneapolis Duration: 17 Jun 2007 → 22 Jun 2007

Conference

Conference	CVPR '07: IEEE Conference on Computer Vision and Pattern Recognition, 2007
City	Minneapolis
Period	17/06/07 → 22/06/07

Access to Document

10.1109/CVPR.2007.382971

Cite this

@conference{6399b2b0be7943839362126814fb5004,

title = "Learning Local Image Descriptors",

abstract = "In this paper we study interest point descriptors for image matching and 3D reconstruction. We examine the building blocks of descriptor algorithms and evaluate numerous combinations of components. Various published descriptors such as SIFT, GLOH, and Spin images can be cast into our framework. For each candidate algorithm we learn good choices for parameters using a training set consisting of patches from a multi-image 3D reconstruction where accurate ground-truth matches are known. The best descriptors were those with log polar histogramming regions and feature vectors constructed from rectified outputs of steerable quadrature filters. At a 95% detection rate these gave one third of the incorrect matches produced by SIFT.",

author = "Winder, {Simon A J} and Matthew Brown",

year = "2007",

month = jun,

doi = "10.1109/CVPR.2007.382971",

language = "English",

note = "CVPR '07: IEEE Conference on Computer Vision and Pattern Recognition, 2007 ; Conference date: 17-06-2007 Through 22-06-2007",

}

TY - CONF

T1 - Learning Local Image Descriptors

AU - Winder, Simon A J

AU - Brown, Matthew

PY - 2007/6

Y1 - 2007/6

N2 - In this paper we study interest point descriptors for image matching and 3D reconstruction. We examine the building blocks of descriptor algorithms and evaluate numerous combinations of components. Various published descriptors such as SIFT, GLOH, and Spin images can be cast into our framework. For each candidate algorithm we learn good choices for parameters using a training set consisting of patches from a multi-image 3D reconstruction where accurate ground-truth matches are known. The best descriptors were those with log polar histogramming regions and feature vectors constructed from rectified outputs of steerable quadrature filters. At a 95% detection rate these gave one third of the incorrect matches produced by SIFT.

AB - In this paper we study interest point descriptors for image matching and 3D reconstruction. We examine the building blocks of descriptor algorithms and evaluate numerous combinations of components. Various published descriptors such as SIFT, GLOH, and Spin images can be cast into our framework. For each candidate algorithm we learn good choices for parameters using a training set consisting of patches from a multi-image 3D reconstruction where accurate ground-truth matches are known. The best descriptors were those with log polar histogramming regions and feature vectors constructed from rectified outputs of steerable quadrature filters. At a 95% detection rate these gave one third of the incorrect matches produced by SIFT.

UR - http://dx.doi.org/10.1109/CVPR.2007.382971

U2 - 10.1109/CVPR.2007.382971

DO - 10.1109/CVPR.2007.382971

M3 - Paper

T2 - CVPR '07: IEEE Conference on Computer Vision and Pattern Recognition, 2007

Y2 - 17 June 2007 through 22 June 2007

ER -

Learning Local Image Descriptors

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this