Extremal Regions Detection Guided by Maxima of Gradient Magnitude

Mehdi Faraji, Jamshid Shambezadeh, Kamal Nasrollahi, Thomas B. Moeslund

Research output: Contribution to journalJournal articleResearchpeer-review

17 Citations (Scopus)
439 Downloads (Pure)

Abstract

A problem of computer vision applications is to detect regions of interest under different imaging conditions. The
state-of-the-art Maximally Stable Extremal Regions (MSER) detects affine covariant regions by applying all possible thresholds on the
input image, and through three main steps including: 1) making a component tree of extremal regions’ evolution (enumeration), 2)
obtaining region stability criterion, and 3) cleaning up. MSER performs very well, but, it does not consider any information about the
boundaries of the regions which are important for detecting repeatable extremal regions. We have shown in this paper that employing
prior information about boundaries of regions results in a novel region detector algorithm that not only outperforms MSER, but avoids
the MSER’s rather complicated steps of enumeration and the cleaning up. To employ the information about the region boundaries we
introduce Maxima of Gradient Magnitudes (MGMs) which are shown to be points that are mostly around the boundaries of the regions.
Having found the MGMs, the method obtains a Global Criterion (GC) for each level of the input image which is used to find Extremum
Levels (ELs). The found ELs are then used to detect extremal regions. The proposed algorithm which is called Extremal Regions of
Extremum Levels (EREL) has been tested on the public benchmark dataset of Mikolajczyk [1]. The obtained experimental results show
that the proposed EREL method outperforms the state-of-the-art methods in terms of the accuracy of the repeatable detected regions
Original languageEnglish
JournalI E E E Transactions on Image Processing
Volume24
Issue number12
Pages (from-to)5401-5415
Number of pages15
ISSN1057-7149
DOIs
Publication statusPublished - 2015

Fingerprint

Cleaning
Stability criteria
Computer vision
Detectors
Imaging techniques

Keywords

  • Extremal Regions of Extremum Levels (EREL)
  • Maxima of Gradient Magnitude (MGM),
  • Maximally Stable Extremal Region (MSER)
  • Feature Detection

Cite this

@article{fb20f887462b4662bb6e4486963952e2,
title = "Extremal Regions Detection Guided by Maxima of Gradient Magnitude",
abstract = "A problem of computer vision applications is to detect regions of interest under different imaging conditions. Thestate-of-the-art Maximally Stable Extremal Regions (MSER) detects affine covariant regions by applying all possible thresholds on theinput image, and through three main steps including: 1) making a component tree of extremal regions’ evolution (enumeration), 2)obtaining region stability criterion, and 3) cleaning up. MSER performs very well, but, it does not consider any information about theboundaries of the regions which are important for detecting repeatable extremal regions. We have shown in this paper that employingprior information about boundaries of regions results in a novel region detector algorithm that not only outperforms MSER, but avoidsthe MSER’s rather complicated steps of enumeration and the cleaning up. To employ the information about the region boundaries weintroduce Maxima of Gradient Magnitudes (MGMs) which are shown to be points that are mostly around the boundaries of the regions.Having found the MGMs, the method obtains a Global Criterion (GC) for each level of the input image which is used to find ExtremumLevels (ELs). The found ELs are then used to detect extremal regions. The proposed algorithm which is called Extremal Regions ofExtremum Levels (EREL) has been tested on the public benchmark dataset of Mikolajczyk [1]. The obtained experimental results showthat the proposed EREL method outperforms the state-of-the-art methods in terms of the accuracy of the repeatable detected regions",
keywords = "Extremal Regions of Extremum Levels (EREL), Maxima of Gradient Magnitude (MGM),, Maximally Stable Extremal Region (MSER), Feature Detection",
author = "Mehdi Faraji and Jamshid Shambezadeh and Kamal Nasrollahi and Moeslund, {Thomas B.}",
year = "2015",
doi = "10.1109/TIP.2015.2477215",
language = "English",
volume = "24",
pages = "5401--5415",
journal = "I E E E Transactions on Image Processing",
issn = "1057-7149",
publisher = "IEEE",
number = "12",

}

Extremal Regions Detection Guided by Maxima of Gradient Magnitude. / Faraji, Mehdi; Shambezadeh, Jamshid; Nasrollahi, Kamal; Moeslund, Thomas B.

In: I E E E Transactions on Image Processing, Vol. 24, No. 12, 2015, p. 5401-5415.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - Extremal Regions Detection Guided by Maxima of Gradient Magnitude

AU - Faraji, Mehdi

AU - Shambezadeh, Jamshid

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2015

Y1 - 2015

N2 - A problem of computer vision applications is to detect regions of interest under different imaging conditions. Thestate-of-the-art Maximally Stable Extremal Regions (MSER) detects affine covariant regions by applying all possible thresholds on theinput image, and through three main steps including: 1) making a component tree of extremal regions’ evolution (enumeration), 2)obtaining region stability criterion, and 3) cleaning up. MSER performs very well, but, it does not consider any information about theboundaries of the regions which are important for detecting repeatable extremal regions. We have shown in this paper that employingprior information about boundaries of regions results in a novel region detector algorithm that not only outperforms MSER, but avoidsthe MSER’s rather complicated steps of enumeration and the cleaning up. To employ the information about the region boundaries weintroduce Maxima of Gradient Magnitudes (MGMs) which are shown to be points that are mostly around the boundaries of the regions.Having found the MGMs, the method obtains a Global Criterion (GC) for each level of the input image which is used to find ExtremumLevels (ELs). The found ELs are then used to detect extremal regions. The proposed algorithm which is called Extremal Regions ofExtremum Levels (EREL) has been tested on the public benchmark dataset of Mikolajczyk [1]. The obtained experimental results showthat the proposed EREL method outperforms the state-of-the-art methods in terms of the accuracy of the repeatable detected regions

AB - A problem of computer vision applications is to detect regions of interest under different imaging conditions. Thestate-of-the-art Maximally Stable Extremal Regions (MSER) detects affine covariant regions by applying all possible thresholds on theinput image, and through three main steps including: 1) making a component tree of extremal regions’ evolution (enumeration), 2)obtaining region stability criterion, and 3) cleaning up. MSER performs very well, but, it does not consider any information about theboundaries of the regions which are important for detecting repeatable extremal regions. We have shown in this paper that employingprior information about boundaries of regions results in a novel region detector algorithm that not only outperforms MSER, but avoidsthe MSER’s rather complicated steps of enumeration and the cleaning up. To employ the information about the region boundaries weintroduce Maxima of Gradient Magnitudes (MGMs) which are shown to be points that are mostly around the boundaries of the regions.Having found the MGMs, the method obtains a Global Criterion (GC) for each level of the input image which is used to find ExtremumLevels (ELs). The found ELs are then used to detect extremal regions. The proposed algorithm which is called Extremal Regions ofExtremum Levels (EREL) has been tested on the public benchmark dataset of Mikolajczyk [1]. The obtained experimental results showthat the proposed EREL method outperforms the state-of-the-art methods in terms of the accuracy of the repeatable detected regions

KW - Extremal Regions of Extremum Levels (EREL)

KW - Maxima of Gradient Magnitude (MGM),

KW - Maximally Stable Extremal Region (MSER)

KW - Feature Detection

U2 - 10.1109/TIP.2015.2477215

DO - 10.1109/TIP.2015.2477215

M3 - Journal article

VL - 24

SP - 5401

EP - 5415

JO - I E E E Transactions on Image Processing

JF - I E E E Transactions on Image Processing

SN - 1057-7149

IS - 12

ER -