Rethinking the prior model for stereo

Hiroshi Ishikawa, Davi Geiger

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Sometimes called the smoothing assumption, the prior model of a stereo matching algorithm is the algorithm's expectation on the surfaces in the world. Any stereo algorithm makes assumptions about the probability to see each surface that can be represented in its representation system. Although the past decade has seen much continued progress in stereo matching algorithms, the prior models used in them have not changed much in three decades: most algorithms still use a smoothing prior that minimizes some function of the difference of depths between neighboring sites, sometimes allowing for discontinuities. However, one system seems to use a very different prior model from all other systems: the human vision system. In this paper, we first report the observations we made in examining human disparity interpolation using stereo pairs with sparse identifiable features. Then we mathematically analyze the implication of using current prior models and explain why the human system seems to use a model that is not only different but in a sense diametrically opposite from all current models. Finally, we propose two candidate models that reflect the behavior of human vision. Although the two models look very different, we show that they are closely related.

Original languageEnglish (US)
Title of host publicationComputer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings
Pages526-537
Number of pages12
Volume3953 LNCS
DOIs
StatePublished - 2006
Event9th European Conference on Computer Vision, ECCV 2006 - Graz, Austria
Duration: May 7 2006May 13 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3953 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other9th European Conference on Computer Vision, ECCV 2006
CountryAustria
CityGraz
Period5/7/065/13/06

Fingerprint

Human Vision
Stereo Matching
Matching Algorithm
Model
Smoothing
Vision System
Discontinuity
Interpolation
Interpolate
Minimise
Human
Observation

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

Ishikawa, H., & Geiger, D. (2006). Rethinking the prior model for stereo. In Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings (Vol. 3953 LNCS, pp. 526-537). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3953 LNCS). https://doi.org/10.1007/11744078_41

Rethinking the prior model for stereo. / Ishikawa, Hiroshi; Geiger, Davi.

Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings. Vol. 3953 LNCS 2006. p. 526-537 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3953 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ishikawa, H & Geiger, D 2006, Rethinking the prior model for stereo. in Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings. vol. 3953 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 3953 LNCS, pp. 526-537, 9th European Conference on Computer Vision, ECCV 2006, Graz, Austria, 5/7/06. https://doi.org/10.1007/11744078_41
Ishikawa H, Geiger D. Rethinking the prior model for stereo. In Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings. Vol. 3953 LNCS. 2006. p. 526-537. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/11744078_41
Ishikawa, Hiroshi ; Geiger, Davi. / Rethinking the prior model for stereo. Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings. Vol. 3953 LNCS 2006. pp. 526-537 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{a898ae613414491488a285d3c04fd5ba,
title = "Rethinking the prior model for stereo",
abstract = "Sometimes called the smoothing assumption, the prior model of a stereo matching algorithm is the algorithm's expectation on the surfaces in the world. Any stereo algorithm makes assumptions about the probability to see each surface that can be represented in its representation system. Although the past decade has seen much continued progress in stereo matching algorithms, the prior models used in them have not changed much in three decades: most algorithms still use a smoothing prior that minimizes some function of the difference of depths between neighboring sites, sometimes allowing for discontinuities. However, one system seems to use a very different prior model from all other systems: the human vision system. In this paper, we first report the observations we made in examining human disparity interpolation using stereo pairs with sparse identifiable features. Then we mathematically analyze the implication of using current prior models and explain why the human system seems to use a model that is not only different but in a sense diametrically opposite from all current models. Finally, we propose two candidate models that reflect the behavior of human vision. Although the two models look very different, we show that they are closely related.",
author = "Hiroshi Ishikawa and Davi Geiger",
year = "2006",
doi = "10.1007/11744078_41",
language = "English (US)",
isbn = "3540338365",
volume = "3953 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "526--537",
booktitle = "Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings",

}

TY - GEN

T1 - Rethinking the prior model for stereo

AU - Ishikawa, Hiroshi

AU - Geiger, Davi

PY - 2006

Y1 - 2006

N2 - Sometimes called the smoothing assumption, the prior model of a stereo matching algorithm is the algorithm's expectation on the surfaces in the world. Any stereo algorithm makes assumptions about the probability to see each surface that can be represented in its representation system. Although the past decade has seen much continued progress in stereo matching algorithms, the prior models used in them have not changed much in three decades: most algorithms still use a smoothing prior that minimizes some function of the difference of depths between neighboring sites, sometimes allowing for discontinuities. However, one system seems to use a very different prior model from all other systems: the human vision system. In this paper, we first report the observations we made in examining human disparity interpolation using stereo pairs with sparse identifiable features. Then we mathematically analyze the implication of using current prior models and explain why the human system seems to use a model that is not only different but in a sense diametrically opposite from all current models. Finally, we propose two candidate models that reflect the behavior of human vision. Although the two models look very different, we show that they are closely related.

AB - Sometimes called the smoothing assumption, the prior model of a stereo matching algorithm is the algorithm's expectation on the surfaces in the world. Any stereo algorithm makes assumptions about the probability to see each surface that can be represented in its representation system. Although the past decade has seen much continued progress in stereo matching algorithms, the prior models used in them have not changed much in three decades: most algorithms still use a smoothing prior that minimizes some function of the difference of depths between neighboring sites, sometimes allowing for discontinuities. However, one system seems to use a very different prior model from all other systems: the human vision system. In this paper, we first report the observations we made in examining human disparity interpolation using stereo pairs with sparse identifiable features. Then we mathematically analyze the implication of using current prior models and explain why the human system seems to use a model that is not only different but in a sense diametrically opposite from all current models. Finally, we propose two candidate models that reflect the behavior of human vision. Although the two models look very different, we show that they are closely related.

UR - http://www.scopus.com/inward/record.url?scp=33745832333&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33745832333&partnerID=8YFLogxK

U2 - 10.1007/11744078_41

DO - 10.1007/11744078_41

M3 - Conference contribution

AN - SCOPUS:33745832333

SN - 3540338365

SN - 9783540338369

VL - 3953 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 526

EP - 537

BT - Computer Vision - ECCV 2006, 9th European Conference on Computer Vision, Proceedings

ER -