Simultaneous Fusion Moves for 3D-label Stereo

Johannes Ulén; Carl Olsson

doi:10.1007/978-3-642-40395-8_7

Simultaneous Fusion Moves for 3D-label Stereo

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

Abstract

Second derivative regularization methods for dense stereo matching is
a topic of intense research. Some of the most successful recent methods employ so called binary fusion moves where the combination of two proposal solutions is computed. In many cases the fusion move can be solved optimally, but the approach is limited to fusing pairs of proposals in each move. For multiple proposals iterative binary fusion may potentially lead to local minima.
In this paper we demonstrate how to simultaneously fuse more than two proposals at the same time for a 2nd order stereo regularizer. The optimization is made possible by effectively computing a generalized distance transform. This allows for computation of messages in linear time in the number of proposals. In addition the approach provides a lower bound on the globally optimal solution of the multi-fusion problem. We verify experimentally that the lower bound is very close to the computed solution, thus providing a near optimal solution.

Original language	English
Title of host publication	Lecture Notes in Computer Science Vol. 8081 (Energy Minimization Methods in Computer Vision and Pattern Recognition)
Editors	Anders Heyden, Fredrik Kahl, Carl Olsson, Magnus Oskarsson, Xue-Cheng Tai
Publisher	Springer
Pages	80-93
Number of pages	14
Volume	8081
ISBN (Print)	978-3-642-40394-1 (print), 978-3-642-40395-8 (online)
DOIs	https://doi.org/10.1007/978-3-642-40395-8_7
Publication status	Published - 2013
Event	Energy Minimization Methods in Computer Vision and Pattern Recognition, 9th International Conference, EMMCVPR 2013 Lund, Sweden, August 19-21, 2013 - Lund, Sweden Duration: 2013 Aug 19 → 2013 Aug 21

Publication series

Name
Volume	8081
ISSN (Print)	1611-3349
ISSN (Electronic)	0302-9743

Conference

Conference	Energy Minimization Methods in Computer Vision and Pattern Recognition, 9th International Conference, EMMCVPR 2013 Lund, Sweden, August 19-21, 2013
Country/Territory	Sweden
City	Lund
Period	2013/08/19 → 2013/08/21

Subject classification (UKÄ)

Computer graphics and computer vision
Mathematical Sciences

Access to Document

10.1007/978-3-642-40395-8_7

Cite this

@inproceedings{060f261a928144349554a06ce1403520,

title = "Simultaneous Fusion Moves for 3D-label Stereo",

abstract = "Second derivative regularization methods for dense stereo matching is a topic of intense research. Some of the most successful recent methods employ so called binary fusion moves where the combination of two proposal solutions is computed. In many cases the fusion move can be solved optimally, but the approach is limited to fusing pairs of proposals in each move. For multiple proposals iterative binary fusion may potentially lead to local minima. In this paper we demonstrate how to simultaneously fuse more than two proposals at the same time for a 2nd order stereo regularizer. The optimization is made possible by effectively computing a generalized distance transform. This allows for computation of messages in linear time in the number of proposals. In addition the approach provides a lower bound on the globally optimal solution of the multi-fusion problem. We verify experimentally that the lower bound is very close to the computed solution, thus providing a near optimal solution.",

author = "Johannes Ul{\'e}n and Carl Olsson",

year = "2013",

doi = "10.1007/978-3-642-40395-8_7",

language = "English",

isbn = "978-3-642-40394-1 (print)",

volume = "8081",

publisher = "Springer",

pages = "80--93",

editor = "Anders Heyden and Fredrik Kahl and Carl Olsson and Magnus Oskarsson and Xue-Cheng Tai",

booktitle = "Lecture Notes in Computer Science Vol. 8081 (Energy Minimization Methods in Computer Vision and Pattern Recognition)",

address = "Germany",

note = "Energy Minimization Methods in Computer Vision and Pattern Recognition, 9th International Conference, EMMCVPR 2013 Lund, Sweden, August 19-21, 2013 ; Conference date: 19-08-2013 Through 21-08-2013",

}

Ulén, J & Olsson, C 2013, Simultaneous Fusion Moves for 3D-label Stereo. in A Heyden, F Kahl, C Olsson, M Oskarsson & X-C Tai (eds), Lecture Notes in Computer Science Vol. 8081 (Energy Minimization Methods in Computer Vision and Pattern Recognition). vol. 8081, Springer, pp. 80-93, Energy Minimization Methods in Computer Vision and Pattern Recognition, 9th International Conference, EMMCVPR 2013 Lund, Sweden, August 19-21, 2013, Lund, Sweden, 2013/08/19. https://doi.org/10.1007/978-3-642-40395-8_7

Simultaneous Fusion Moves for 3D-label Stereo. / Ulén, Johannes; Olsson, Carl.
Lecture Notes in Computer Science Vol. 8081 (Energy Minimization Methods in Computer Vision and Pattern Recognition). ed. / Anders Heyden; Fredrik Kahl; Carl Olsson; Magnus Oskarsson; Xue-Cheng Tai. Vol. 8081 Springer, 2013. p. 80-93.

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

TY - GEN

T1 - Simultaneous Fusion Moves for 3D-label Stereo

AU - Ulén, Johannes

AU - Olsson, Carl

PY - 2013

Y1 - 2013

N2 - Second derivative regularization methods for dense stereo matching is a topic of intense research. Some of the most successful recent methods employ so called binary fusion moves where the combination of two proposal solutions is computed. In many cases the fusion move can be solved optimally, but the approach is limited to fusing pairs of proposals in each move. For multiple proposals iterative binary fusion may potentially lead to local minima. In this paper we demonstrate how to simultaneously fuse more than two proposals at the same time for a 2nd order stereo regularizer. The optimization is made possible by effectively computing a generalized distance transform. This allows for computation of messages in linear time in the number of proposals. In addition the approach provides a lower bound on the globally optimal solution of the multi-fusion problem. We verify experimentally that the lower bound is very close to the computed solution, thus providing a near optimal solution.

AB - Second derivative regularization methods for dense stereo matching is a topic of intense research. Some of the most successful recent methods employ so called binary fusion moves where the combination of two proposal solutions is computed. In many cases the fusion move can be solved optimally, but the approach is limited to fusing pairs of proposals in each move. For multiple proposals iterative binary fusion may potentially lead to local minima. In this paper we demonstrate how to simultaneously fuse more than two proposals at the same time for a 2nd order stereo regularizer. The optimization is made possible by effectively computing a generalized distance transform. This allows for computation of messages in linear time in the number of proposals. In addition the approach provides a lower bound on the globally optimal solution of the multi-fusion problem. We verify experimentally that the lower bound is very close to the computed solution, thus providing a near optimal solution.

U2 - 10.1007/978-3-642-40395-8_7

DO - 10.1007/978-3-642-40395-8_7

M3 - Paper in conference proceeding

SN - 978-3-642-40394-1 (print)

SN - 978-3-642-40395-8 (online)

VL - 8081

SP - 80

EP - 93

BT - Lecture Notes in Computer Science Vol. 8081 (Energy Minimization Methods in Computer Vision and Pattern Recognition)

A2 - Heyden, Anders

A2 - Kahl, Fredrik

A2 - Olsson, Carl

A2 - Oskarsson, Magnus

A2 - Tai, Xue-Cheng

PB - Springer

T2 - Energy Minimization Methods in Computer Vision and Pattern Recognition, 9th International Conference, EMMCVPR 2013 Lund, Sweden, August 19-21, 2013

Y2 - 19 August 2013 through 21 August 2013

ER -

Simultaneous Fusion Moves for 3D-label Stereo

Abstract

Publication series

Conference

Subject classification (UKÄ)

Access to Document

Fingerprint

Cite this