Relaxation methods for supervised image segmentation

Michael W. Hansen, William Evan Higgins

Research output: Contribution to journalArticle

32 Citations (Scopus)

Abstract

We propose two methods for supervised image segmentation: supervised relaxation labeling and watershed-driven relaxation labeling. The methods are particularly well suited to problems in 3D medical image analysis, where the images are large, the regions are topologically complex, and the tolerance of errors is low. Each method uses predefined cues for supervision. The cues can be defined interactively or automatically, depending on the application. The cues provide statistical region information and region topological constraints. Supervised relaxation labeling exhibits strong noise resilience. Watershed-driven relaxation labeling combines the strengths of watershed analysis and supervised relaxation labeling to give a computationally efficient noise-resistant method. Extensive results for 2D and 3D images illustrate the effectiveness of the methods.

Original languageEnglish (US)
Pages (from-to)949-962
Number of pages14
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume19
Issue number9
DOIs
StatePublished - Dec 1 1997

Fingerprint

Relaxation Method
Image segmentation
Image Segmentation
Labeling
Watersheds
3D Image
Medical Image Analysis
Resilience
Image analysis
Tolerance

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Vision and Pattern Recognition
  • Computational Theory and Mathematics
  • Artificial Intelligence
  • Applied Mathematics

Cite this

@article{d38a0697c35549058a13d74393c5bfea,
title = "Relaxation methods for supervised image segmentation",
abstract = "We propose two methods for supervised image segmentation: supervised relaxation labeling and watershed-driven relaxation labeling. The methods are particularly well suited to problems in 3D medical image analysis, where the images are large, the regions are topologically complex, and the tolerance of errors is low. Each method uses predefined cues for supervision. The cues can be defined interactively or automatically, depending on the application. The cues provide statistical region information and region topological constraints. Supervised relaxation labeling exhibits strong noise resilience. Watershed-driven relaxation labeling combines the strengths of watershed analysis and supervised relaxation labeling to give a computationally efficient noise-resistant method. Extensive results for 2D and 3D images illustrate the effectiveness of the methods.",
author = "Hansen, {Michael W.} and Higgins, {William Evan}",
year = "1997",
month = "12",
day = "1",
doi = "10.1109/34.615445",
language = "English (US)",
volume = "19",
pages = "949--962",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "9",

}

Relaxation methods for supervised image segmentation. / Hansen, Michael W.; Higgins, William Evan.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 19, No. 9, 01.12.1997, p. 949-962.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Relaxation methods for supervised image segmentation

AU - Hansen, Michael W.

AU - Higgins, William Evan

PY - 1997/12/1

Y1 - 1997/12/1

N2 - We propose two methods for supervised image segmentation: supervised relaxation labeling and watershed-driven relaxation labeling. The methods are particularly well suited to problems in 3D medical image analysis, where the images are large, the regions are topologically complex, and the tolerance of errors is low. Each method uses predefined cues for supervision. The cues can be defined interactively or automatically, depending on the application. The cues provide statistical region information and region topological constraints. Supervised relaxation labeling exhibits strong noise resilience. Watershed-driven relaxation labeling combines the strengths of watershed analysis and supervised relaxation labeling to give a computationally efficient noise-resistant method. Extensive results for 2D and 3D images illustrate the effectiveness of the methods.

AB - We propose two methods for supervised image segmentation: supervised relaxation labeling and watershed-driven relaxation labeling. The methods are particularly well suited to problems in 3D medical image analysis, where the images are large, the regions are topologically complex, and the tolerance of errors is low. Each method uses predefined cues for supervision. The cues can be defined interactively or automatically, depending on the application. The cues provide statistical region information and region topological constraints. Supervised relaxation labeling exhibits strong noise resilience. Watershed-driven relaxation labeling combines the strengths of watershed analysis and supervised relaxation labeling to give a computationally efficient noise-resistant method. Extensive results for 2D and 3D images illustrate the effectiveness of the methods.

UR - http://www.scopus.com/inward/record.url?scp=0031234988&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031234988&partnerID=8YFLogxK

U2 - 10.1109/34.615445

DO - 10.1109/34.615445

M3 - Article

AN - SCOPUS:0031234988

VL - 19

SP - 949

EP - 962

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 9

ER -