Loading [MathJax]/extensions/tex2jax.js

Localization Accuracy of Interest Point Detectors with Different Scale Space Representations

Publikation: Beitrag in Buch/Bericht/Sammelwerk/KonferenzbandAufsatz in KonferenzbandForschungPeer-Review

Autorschaft

Plum Print visual indicator of research metrics
  • Citations
    • Citation Indexes: 3
  • Captures
    • Readers: 9
see details

Details

OriginalspracheEnglisch
Titel des Sammelwerks11th IEEE International Conference on Advanced Video and Signal-Based Surveillance
UntertitelAVSS 2014
Herausgeber (Verlag)Institute of Electrical and Electronics Engineers Inc.
Seiten247-252
Seitenumfang6
ISBN (elektronisch)9781479948710
PublikationsstatusVeröffentlicht - 8 Okt. 2014
Veranstaltung11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014 - Seoul, Südkorea
Dauer: 26 Aug. 201429 Aug. 2014
Konferenznummer: 11

Publikationsreihe

Name11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014

Abstract

The detection of scale invariant image features is a fundamental task for computer vision applications like object recognition or re-identification. Features are localized by computing extrema of the gradients in the Laplacian of Gaussian (LoG) scale space. The most popular detector for scale invariant features is the SIFT detector which uses the Difference of Gaussians (DoG) pyramid as an approximation of the LoG. Recently, the alternative interest point (ALP) detector demonstrated its strength in fast computation on highly parallel architectures like the GPU. It uses the LoG scale space representation for the localization of interest points. This paper evaluates the localization accuracy of ALP in comparison to SIFT. By using synthetic images, it is demonstrated that both localization approaches show a systematic error which is dependent on the subpixel position of the feature. The error increases with the scale of the detected feature. However, using the LoG instead of the DoG representation reduces the maximum systematic error by 77 %. For the evaluation with natural images, benchmark data sets are used. The repeatability criterion evaluates the accuracy of the detectors. The LoG based detector results in up to 16 % higher repeatability. The comparisons are completed with a reference feature localization which uses a signal based approach for the gradient approximation. Based on this approach, a new feature selection criterion is proposed.

ASJC Scopus Sachgebiete

Zitieren

Localization Accuracy of Interest Point Detectors with Different Scale Space Representations. / Cordes, Kai; Rosenhahn, Bodo; Ostermann, Jörn.
11th IEEE International Conference on Advanced Video and Signal-Based Surveillance: AVSS 2014. Institute of Electrical and Electronics Engineers Inc., 2014. S. 247-252 6918676 (11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014).

Publikation: Beitrag in Buch/Bericht/Sammelwerk/KonferenzbandAufsatz in KonferenzbandForschungPeer-Review

Cordes, K, Rosenhahn, B & Ostermann, J 2014, Localization Accuracy of Interest Point Detectors with Different Scale Space Representations. in 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance: AVSS 2014., 6918676, 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014, Institute of Electrical and Electronics Engineers Inc., S. 247-252, 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014, Seoul, Südkorea, 26 Aug. 2014. https://doi.org/10.1109/AVSS.2014.6918676
Cordes, K., Rosenhahn, B., & Ostermann, J. (2014). Localization Accuracy of Interest Point Detectors with Different Scale Space Representations. In 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance: AVSS 2014 (S. 247-252). Artikel 6918676 (11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/AVSS.2014.6918676
Cordes K, Rosenhahn B, Ostermann J. Localization Accuracy of Interest Point Detectors with Different Scale Space Representations. in 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance: AVSS 2014. Institute of Electrical and Electronics Engineers Inc. 2014. S. 247-252. 6918676. (11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014). doi: 10.1109/AVSS.2014.6918676
Cordes, Kai ; Rosenhahn, Bodo ; Ostermann, Jörn. / Localization Accuracy of Interest Point Detectors with Different Scale Space Representations. 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance: AVSS 2014. Institute of Electrical and Electronics Engineers Inc., 2014. S. 247-252 (11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014).
Download
@inproceedings{74dc43313c1f49fb99b59bf5b936551c,
title = "Localization Accuracy of Interest Point Detectors with Different Scale Space Representations",
abstract = "The detection of scale invariant image features is a fundamental task for computer vision applications like object recognition or re-identification. Features are localized by computing extrema of the gradients in the Laplacian of Gaussian (LoG) scale space. The most popular detector for scale invariant features is the SIFT detector which uses the Difference of Gaussians (DoG) pyramid as an approximation of the LoG. Recently, the alternative interest point (ALP) detector demonstrated its strength in fast computation on highly parallel architectures like the GPU. It uses the LoG scale space representation for the localization of interest points. This paper evaluates the localization accuracy of ALP in comparison to SIFT. By using synthetic images, it is demonstrated that both localization approaches show a systematic error which is dependent on the subpixel position of the feature. The error increases with the scale of the detected feature. However, using the LoG instead of the DoG representation reduces the maximum systematic error by 77 %. For the evaluation with natural images, benchmark data sets are used. The repeatability criterion evaluates the accuracy of the detectors. The LoG based detector results in up to 16 % higher repeatability. The comparisons are completed with a reference feature localization which uses a signal based approach for the gradient approximation. Based on this approach, a new feature selection criterion is proposed.",
author = "Kai Cordes and Bodo Rosenhahn and J{\"o}rn Ostermann",
year = "2014",
month = oct,
day = "8",
doi = "10.1109/AVSS.2014.6918676",
language = "English",
series = "11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "247--252",
booktitle = "11th IEEE International Conference on Advanced Video and Signal-Based Surveillance",
address = "United States",
note = "11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014 ; Conference date: 26-08-2014 Through 29-08-2014",

}

Download

TY - GEN

T1 - Localization Accuracy of Interest Point Detectors with Different Scale Space Representations

AU - Cordes, Kai

AU - Rosenhahn, Bodo

AU - Ostermann, Jörn

N1 - Conference code: 11

PY - 2014/10/8

Y1 - 2014/10/8

N2 - The detection of scale invariant image features is a fundamental task for computer vision applications like object recognition or re-identification. Features are localized by computing extrema of the gradients in the Laplacian of Gaussian (LoG) scale space. The most popular detector for scale invariant features is the SIFT detector which uses the Difference of Gaussians (DoG) pyramid as an approximation of the LoG. Recently, the alternative interest point (ALP) detector demonstrated its strength in fast computation on highly parallel architectures like the GPU. It uses the LoG scale space representation for the localization of interest points. This paper evaluates the localization accuracy of ALP in comparison to SIFT. By using synthetic images, it is demonstrated that both localization approaches show a systematic error which is dependent on the subpixel position of the feature. The error increases with the scale of the detected feature. However, using the LoG instead of the DoG representation reduces the maximum systematic error by 77 %. For the evaluation with natural images, benchmark data sets are used. The repeatability criterion evaluates the accuracy of the detectors. The LoG based detector results in up to 16 % higher repeatability. The comparisons are completed with a reference feature localization which uses a signal based approach for the gradient approximation. Based on this approach, a new feature selection criterion is proposed.

AB - The detection of scale invariant image features is a fundamental task for computer vision applications like object recognition or re-identification. Features are localized by computing extrema of the gradients in the Laplacian of Gaussian (LoG) scale space. The most popular detector for scale invariant features is the SIFT detector which uses the Difference of Gaussians (DoG) pyramid as an approximation of the LoG. Recently, the alternative interest point (ALP) detector demonstrated its strength in fast computation on highly parallel architectures like the GPU. It uses the LoG scale space representation for the localization of interest points. This paper evaluates the localization accuracy of ALP in comparison to SIFT. By using synthetic images, it is demonstrated that both localization approaches show a systematic error which is dependent on the subpixel position of the feature. The error increases with the scale of the detected feature. However, using the LoG instead of the DoG representation reduces the maximum systematic error by 77 %. For the evaluation with natural images, benchmark data sets are used. The repeatability criterion evaluates the accuracy of the detectors. The LoG based detector results in up to 16 % higher repeatability. The comparisons are completed with a reference feature localization which uses a signal based approach for the gradient approximation. Based on this approach, a new feature selection criterion is proposed.

UR - http://www.scopus.com/inward/record.url?scp=84909954483&partnerID=8YFLogxK

U2 - 10.1109/AVSS.2014.6918676

DO - 10.1109/AVSS.2014.6918676

M3 - Conference contribution

AN - SCOPUS:84909954483

T3 - 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014

SP - 247

EP - 252

BT - 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 11th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS 2014

Y2 - 26 August 2014 through 29 August 2014

ER -

Von denselben Autoren