Total variation on a tree

Vladimir Kolmogorov, Thomas Pock, Michal Rolinek

Research output: Contribution to journalArticleResearchpeer-review

Abstract

We consider the problem of minimizing the continuous valued total variation subject to different unary terms on trees and propose fast direct algorithms based on dynamic programming to solve these problems. We treat both the convex and the nonconvex case and derive worst-case complexities that are equal to or better than existing methods. We show applications to total variation based two dimensional image processing and computer vision problems based on a Lagrangian decomposition approach. The resulting algorithms are very effcient, offer a high degree of parallelism, and come along with memory requirements which are only in the order of the number of image pixels.

Original languageEnglish
Pages (from-to)605-636
Number of pages32
JournalSIAM Journal on Imaging Sciences
Volume9
Issue number2
DOIs
Publication statusPublished - 3 May 2016

Fingerprint

Total Variation
Dynamic programming
Computer vision
Image processing
Pixels
Unary
Decomposition
Data storage equipment
Computer Vision
Parallelism
Dynamic Programming
Image Processing
Pixel
Decompose
Requirements
Term

Keywords

  • Dynamic programming
  • Primal-dual optimization
  • Total variation

ASJC Scopus subject areas

  • Mathematics(all)
  • Applied Mathematics

Cite this

Total variation on a tree. / Kolmogorov, Vladimir; Pock, Thomas; Rolinek, Michal.

In: SIAM Journal on Imaging Sciences, Vol. 9, No. 2, 03.05.2016, p. 605-636.

Research output: Contribution to journalArticleResearchpeer-review

Kolmogorov, Vladimir ; Pock, Thomas ; Rolinek, Michal. / Total variation on a tree. In: SIAM Journal on Imaging Sciences. 2016 ; Vol. 9, No. 2. pp. 605-636.
@article{97f1f708a4434b8f8e413cfac15b0c14,
title = "Total variation on a tree",
abstract = "We consider the problem of minimizing the continuous valued total variation subject to different unary terms on trees and propose fast direct algorithms based on dynamic programming to solve these problems. We treat both the convex and the nonconvex case and derive worst-case complexities that are equal to or better than existing methods. We show applications to total variation based two dimensional image processing and computer vision problems based on a Lagrangian decomposition approach. The resulting algorithms are very effcient, offer a high degree of parallelism, and come along with memory requirements which are only in the order of the number of image pixels.",
keywords = "Dynamic programming, Primal-dual optimization, Total variation",
author = "Vladimir Kolmogorov and Thomas Pock and Michal Rolinek",
year = "2016",
month = "5",
day = "3",
doi = "10.1137/15M1010257",
language = "English",
volume = "9",
pages = "605--636",
journal = "SIAM journal on imaging sciences",
issn = "1936-4954",
publisher = "Society for Industrial and Applied Mathematics Publications",
number = "2",

}

TY - JOUR

T1 - Total variation on a tree

AU - Kolmogorov, Vladimir

AU - Pock, Thomas

AU - Rolinek, Michal

PY - 2016/5/3

Y1 - 2016/5/3

N2 - We consider the problem of minimizing the continuous valued total variation subject to different unary terms on trees and propose fast direct algorithms based on dynamic programming to solve these problems. We treat both the convex and the nonconvex case and derive worst-case complexities that are equal to or better than existing methods. We show applications to total variation based two dimensional image processing and computer vision problems based on a Lagrangian decomposition approach. The resulting algorithms are very effcient, offer a high degree of parallelism, and come along with memory requirements which are only in the order of the number of image pixels.

AB - We consider the problem of minimizing the continuous valued total variation subject to different unary terms on trees and propose fast direct algorithms based on dynamic programming to solve these problems. We treat both the convex and the nonconvex case and derive worst-case complexities that are equal to or better than existing methods. We show applications to total variation based two dimensional image processing and computer vision problems based on a Lagrangian decomposition approach. The resulting algorithms are very effcient, offer a high degree of parallelism, and come along with memory requirements which are only in the order of the number of image pixels.

KW - Dynamic programming

KW - Primal-dual optimization

KW - Total variation

UR - http://www.scopus.com/inward/record.url?scp=84976629684&partnerID=8YFLogxK

U2 - 10.1137/15M1010257

DO - 10.1137/15M1010257

M3 - Article

VL - 9

SP - 605

EP - 636

JO - SIAM journal on imaging sciences

JF - SIAM journal on imaging sciences

SN - 1936-4954

IS - 2

ER -