Error measures for objective assessment of scene segmentation algorithms

W. A. Yasnoff, W. Galbraith, J. W. Bacus

Research output: Contribution to journalArticle

Abstract

Scene segmentation is an important element in pattern recognition problems. Previous efforts to evaluate and compare scene segmentation procedures have been largely subjective. Quantitative error measures would facilitate objective comparison of scene segmentation algorithms. A theoretical discussion leading to a new generalized quantitative error measure, G2, based on comparison of both pixel class proportions and spatial distributions of 'true' and test segmentation, is presented. This error measure was tested on 14 manual segmentations and 40 gynecologic cytology specimens segmented with five different scene segmentation techniques. Results indicate that G2 seems to have the desirable properties of correlation with human observation, categorization of error allowing for weighting, invariance with picture size and ease of computation necessary for a useful scene segmentation error measure.

Original languageEnglish (US)
Pages (from-to)107-121
Number of pages15
JournalAnalytical and Quantitative Cytology
Volume1
Issue number2
StatePublished - 1979
Externally publishedYes

Fingerprint

Cell Biology
Observation

ASJC Scopus subject areas

  • Anatomy
  • Histology

Cite this

Yasnoff, W. A., Galbraith, W., & Bacus, J. W. (1979). Error measures for objective assessment of scene segmentation algorithms. Analytical and Quantitative Cytology, 1(2), 107-121.

Error measures for objective assessment of scene segmentation algorithms. / Yasnoff, W. A.; Galbraith, W.; Bacus, J. W.

In: Analytical and Quantitative Cytology, Vol. 1, No. 2, 1979, p. 107-121.

Research output: Contribution to journalArticle

Yasnoff, WA, Galbraith, W & Bacus, JW 1979, 'Error measures for objective assessment of scene segmentation algorithms', Analytical and Quantitative Cytology, vol. 1, no. 2, pp. 107-121.
Yasnoff, W. A. ; Galbraith, W. ; Bacus, J. W. / Error measures for objective assessment of scene segmentation algorithms. In: Analytical and Quantitative Cytology. 1979 ; Vol. 1, No. 2. pp. 107-121.
@article{5c822f42d21a42e3833f5832b93b319f,
title = "Error measures for objective assessment of scene segmentation algorithms",
abstract = "Scene segmentation is an important element in pattern recognition problems. Previous efforts to evaluate and compare scene segmentation procedures have been largely subjective. Quantitative error measures would facilitate objective comparison of scene segmentation algorithms. A theoretical discussion leading to a new generalized quantitative error measure, G2, based on comparison of both pixel class proportions and spatial distributions of 'true' and test segmentation, is presented. This error measure was tested on 14 manual segmentations and 40 gynecologic cytology specimens segmented with five different scene segmentation techniques. Results indicate that G2 seems to have the desirable properties of correlation with human observation, categorization of error allowing for weighting, invariance with picture size and ease of computation necessary for a useful scene segmentation error measure.",
author = "Yasnoff, {W. A.} and W. Galbraith and Bacus, {J. W.}",
year = "1979",
language = "English (US)",
volume = "1",
pages = "107--121",
journal = "Analytical and Quantitative Cytopathology and Histopathology",
issn = "0301-102X",
publisher = "John Rylands University Library",
number = "2",

}

TY - JOUR

T1 - Error measures for objective assessment of scene segmentation algorithms

AU - Yasnoff, W. A.

AU - Galbraith, W.

AU - Bacus, J. W.

PY - 1979

Y1 - 1979

N2 - Scene segmentation is an important element in pattern recognition problems. Previous efforts to evaluate and compare scene segmentation procedures have been largely subjective. Quantitative error measures would facilitate objective comparison of scene segmentation algorithms. A theoretical discussion leading to a new generalized quantitative error measure, G2, based on comparison of both pixel class proportions and spatial distributions of 'true' and test segmentation, is presented. This error measure was tested on 14 manual segmentations and 40 gynecologic cytology specimens segmented with five different scene segmentation techniques. Results indicate that G2 seems to have the desirable properties of correlation with human observation, categorization of error allowing for weighting, invariance with picture size and ease of computation necessary for a useful scene segmentation error measure.

AB - Scene segmentation is an important element in pattern recognition problems. Previous efforts to evaluate and compare scene segmentation procedures have been largely subjective. Quantitative error measures would facilitate objective comparison of scene segmentation algorithms. A theoretical discussion leading to a new generalized quantitative error measure, G2, based on comparison of both pixel class proportions and spatial distributions of 'true' and test segmentation, is presented. This error measure was tested on 14 manual segmentations and 40 gynecologic cytology specimens segmented with five different scene segmentation techniques. Results indicate that G2 seems to have the desirable properties of correlation with human observation, categorization of error allowing for weighting, invariance with picture size and ease of computation necessary for a useful scene segmentation error measure.

UR - http://www.scopus.com/inward/record.url?scp=0018577059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0018577059&partnerID=8YFLogxK

M3 - Article

C2 - 396835

AN - SCOPUS:0018577059

VL - 1

SP - 107

EP - 121

JO - Analytical and Quantitative Cytopathology and Histopathology

JF - Analytical and Quantitative Cytopathology and Histopathology

SN - 0301-102X

IS - 2

ER -