Multi-vector feature space based on pseudo-Euclidean space and oblique basis

Yasuo Yamane, Tadashi Hoshiai, Hiroshi Tsuda, Manabu Ohta, Kaoru Katayama, Hiroshi Ishikawa

Research output: Contribution to journalArticle

Abstract

The Earth Mover's Distance (EMD) and the quadratic-form distance (QFD) are representative distances used in similarity searches of images. Although the QFD greatly outperforms the EMD in speed, the EMD outperform the QFD in performance. The EMD, however, has almost no theoretical justification and requires high computation costs. We propose a feature space model we call a "multi-vector feature space based on pseudo-Euclidean space and an oblique basis (MVPO)." In MVPO, an object such as an image is represented by a vector set (roughly speaking, a solid) and the EMD is reinterpreted as the distance between vector sets while the QFD is reinterpreted as the distance between the centroids of vector sets. Therefore MVPO gives a common geometrical view to these distances. We hypothesized that in MVPO the entity of an image is represented by a vector set (solid) and geometrical reasoning is applicable to MVPO. Our hypothesis explains well that the EMD outperforms the QFD in performance because the centroid of a solid is the simplest approximation of it. Our hypothesis implies that the performance of the QFD should be good when solids are far apart but bad when they are close together. We conjectured that discriminability would decline-that is, dissimilar images would be judged to be similar-when the centroids of solids are very close. Our experiment supported this conjecture. And from our hypothesis we conjectured that by making an original solid simpler, we can make an approximation method that has better performance than the QFD and faster than the EMD. The results of our experiment with this method supported our conjecture and consequently our hypothesis.

Original languageEnglish
Pages (from-to)287-308
Number of pages22
JournalMultimedia Tools and Applications
Volume31
Issue number3
DOIs
Publication statusPublished - Dec 2006

Fingerprint

Pseudo-Euclidean Space
Oblique
Feature Space
Earth (planet)
Quadratic form
Centroid
Experiments

Keywords

  • Earth mover's distance
  • Feature space
  • Multi-vector feature space
  • Oblique basis
  • Pseudo-Euclidean space
  • Quadratic-form distance
  • Similarity search

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Information Systems
  • Computer Graphics and Computer-Aided Design
  • Software
  • Theoretical Computer Science
  • Computational Theory and Mathematics

Cite this

Multi-vector feature space based on pseudo-Euclidean space and oblique basis. / Yamane, Yasuo; Hoshiai, Tadashi; Tsuda, Hiroshi; Ohta, Manabu; Katayama, Kaoru; Ishikawa, Hiroshi.

In: Multimedia Tools and Applications, Vol. 31, No. 3, 12.2006, p. 287-308.

Research output: Contribution to journalArticle

Yamane, Yasuo ; Hoshiai, Tadashi ; Tsuda, Hiroshi ; Ohta, Manabu ; Katayama, Kaoru ; Ishikawa, Hiroshi. / Multi-vector feature space based on pseudo-Euclidean space and oblique basis. In: Multimedia Tools and Applications. 2006 ; Vol. 31, No. 3. pp. 287-308.
@article{a4840e46da144bcabf9c356d925c8614,
title = "Multi-vector feature space based on pseudo-Euclidean space and oblique basis",
abstract = "The Earth Mover's Distance (EMD) and the quadratic-form distance (QFD) are representative distances used in similarity searches of images. Although the QFD greatly outperforms the EMD in speed, the EMD outperform the QFD in performance. The EMD, however, has almost no theoretical justification and requires high computation costs. We propose a feature space model we call a {"}multi-vector feature space based on pseudo-Euclidean space and an oblique basis (MVPO).{"} In MVPO, an object such as an image is represented by a vector set (roughly speaking, a solid) and the EMD is reinterpreted as the distance between vector sets while the QFD is reinterpreted as the distance between the centroids of vector sets. Therefore MVPO gives a common geometrical view to these distances. We hypothesized that in MVPO the entity of an image is represented by a vector set (solid) and geometrical reasoning is applicable to MVPO. Our hypothesis explains well that the EMD outperforms the QFD in performance because the centroid of a solid is the simplest approximation of it. Our hypothesis implies that the performance of the QFD should be good when solids are far apart but bad when they are close together. We conjectured that discriminability would decline-that is, dissimilar images would be judged to be similar-when the centroids of solids are very close. Our experiment supported this conjecture. And from our hypothesis we conjectured that by making an original solid simpler, we can make an approximation method that has better performance than the QFD and faster than the EMD. The results of our experiment with this method supported our conjecture and consequently our hypothesis.",
keywords = "Earth mover's distance, Feature space, Multi-vector feature space, Oblique basis, Pseudo-Euclidean space, Quadratic-form distance, Similarity search",
author = "Yasuo Yamane and Tadashi Hoshiai and Hiroshi Tsuda and Manabu Ohta and Kaoru Katayama and Hiroshi Ishikawa",
year = "2006",
month = "12",
doi = "10.1007/s11042-006-0045-z",
language = "English",
volume = "31",
pages = "287--308",
journal = "Multimedia Tools and Applications",
issn = "1380-7501",
publisher = "Springer Netherlands",
number = "3",

}

TY - JOUR

T1 - Multi-vector feature space based on pseudo-Euclidean space and oblique basis

AU - Yamane, Yasuo

AU - Hoshiai, Tadashi

AU - Tsuda, Hiroshi

AU - Ohta, Manabu

AU - Katayama, Kaoru

AU - Ishikawa, Hiroshi

PY - 2006/12

Y1 - 2006/12

N2 - The Earth Mover's Distance (EMD) and the quadratic-form distance (QFD) are representative distances used in similarity searches of images. Although the QFD greatly outperforms the EMD in speed, the EMD outperform the QFD in performance. The EMD, however, has almost no theoretical justification and requires high computation costs. We propose a feature space model we call a "multi-vector feature space based on pseudo-Euclidean space and an oblique basis (MVPO)." In MVPO, an object such as an image is represented by a vector set (roughly speaking, a solid) and the EMD is reinterpreted as the distance between vector sets while the QFD is reinterpreted as the distance between the centroids of vector sets. Therefore MVPO gives a common geometrical view to these distances. We hypothesized that in MVPO the entity of an image is represented by a vector set (solid) and geometrical reasoning is applicable to MVPO. Our hypothesis explains well that the EMD outperforms the QFD in performance because the centroid of a solid is the simplest approximation of it. Our hypothesis implies that the performance of the QFD should be good when solids are far apart but bad when they are close together. We conjectured that discriminability would decline-that is, dissimilar images would be judged to be similar-when the centroids of solids are very close. Our experiment supported this conjecture. And from our hypothesis we conjectured that by making an original solid simpler, we can make an approximation method that has better performance than the QFD and faster than the EMD. The results of our experiment with this method supported our conjecture and consequently our hypothesis.

AB - The Earth Mover's Distance (EMD) and the quadratic-form distance (QFD) are representative distances used in similarity searches of images. Although the QFD greatly outperforms the EMD in speed, the EMD outperform the QFD in performance. The EMD, however, has almost no theoretical justification and requires high computation costs. We propose a feature space model we call a "multi-vector feature space based on pseudo-Euclidean space and an oblique basis (MVPO)." In MVPO, an object such as an image is represented by a vector set (roughly speaking, a solid) and the EMD is reinterpreted as the distance between vector sets while the QFD is reinterpreted as the distance between the centroids of vector sets. Therefore MVPO gives a common geometrical view to these distances. We hypothesized that in MVPO the entity of an image is represented by a vector set (solid) and geometrical reasoning is applicable to MVPO. Our hypothesis explains well that the EMD outperforms the QFD in performance because the centroid of a solid is the simplest approximation of it. Our hypothesis implies that the performance of the QFD should be good when solids are far apart but bad when they are close together. We conjectured that discriminability would decline-that is, dissimilar images would be judged to be similar-when the centroids of solids are very close. Our experiment supported this conjecture. And from our hypothesis we conjectured that by making an original solid simpler, we can make an approximation method that has better performance than the QFD and faster than the EMD. The results of our experiment with this method supported our conjecture and consequently our hypothesis.

KW - Earth mover's distance

KW - Feature space

KW - Multi-vector feature space

KW - Oblique basis

KW - Pseudo-Euclidean space

KW - Quadratic-form distance

KW - Similarity search

UR - http://www.scopus.com/inward/record.url?scp=33751088244&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33751088244&partnerID=8YFLogxK

U2 - 10.1007/s11042-006-0045-z

DO - 10.1007/s11042-006-0045-z

M3 - Article

VL - 31

SP - 287

EP - 308

JO - Multimedia Tools and Applications

JF - Multimedia Tools and Applications

SN - 1380-7501

IS - 3

ER -