Robust image recognition by fusion of contextual information

Xubo Song, Yaser Abu-Mostafa, Joseph Sill, Harvey Kasdan, Misha Pavel

Research output: Contribution to journalArticle

20 Citations (Scopus)

Abstract

This paper studies the fusion of contextual information in pattern recognition, with applications to biomedical image identification. In the real world there are cases where the identity of an object is ambiguous if the classification is based only on its own features. It is helpful to reduce the ambiguity by utilizing extra information, referred to as context, provided by accompanying objects. We investigate two techniques that incorporate context. The first approach, based on compound Bayesian theory, incorporates context by fusing the measurements of all objects under consideration. It is an optimal strategy in terms of achieving minimum set-by-set error probability. The second approach fuses the measurements of an object with explicitly extracted context. Its linear computational complexity makes it more tractable than the first approach, which requires exponential computation. These two techniques are applied to two medical applications: white blood cell image classification and microscopic urinalysis. It is demonstrated that superior classification performances are achieved by using context. In our particular applications, it reduces overall classification error, as well as false positive and false negative diagnosis rates.

Original languageEnglish (US)
Pages (from-to)277-287
Number of pages11
JournalInformation Fusion
Volume3
Issue number4
DOIs
StatePublished - Dec 2002

Fingerprint

Image recognition
Fusion reactions
Image classification
Medical applications
Electric fuses
Pattern recognition
Computational complexity
Blood
Cells

Keywords

  • Compound Bayesian Theory
  • Context
  • Contextual information
  • Fusion
  • Image recognition
  • Pattern recognition

ASJC Scopus subject areas

  • Information Systems
  • Signal Processing
  • Engineering(all)

Cite this

Robust image recognition by fusion of contextual information. / Song, Xubo; Abu-Mostafa, Yaser; Sill, Joseph; Kasdan, Harvey; Pavel, Misha.

In: Information Fusion, Vol. 3, No. 4, 12.2002, p. 277-287.

Research output: Contribution to journalArticle

Song, X, Abu-Mostafa, Y, Sill, J, Kasdan, H & Pavel, M 2002, 'Robust image recognition by fusion of contextual information', Information Fusion, vol. 3, no. 4, pp. 277-287. https://doi.org/10.1016/S1566-2535(02)00092-1
Song, Xubo ; Abu-Mostafa, Yaser ; Sill, Joseph ; Kasdan, Harvey ; Pavel, Misha. / Robust image recognition by fusion of contextual information. In: Information Fusion. 2002 ; Vol. 3, No. 4. pp. 277-287.
@article{c19036eb4811481096cd529d992d8290,
title = "Robust image recognition by fusion of contextual information",
abstract = "This paper studies the fusion of contextual information in pattern recognition, with applications to biomedical image identification. In the real world there are cases where the identity of an object is ambiguous if the classification is based only on its own features. It is helpful to reduce the ambiguity by utilizing extra information, referred to as context, provided by accompanying objects. We investigate two techniques that incorporate context. The first approach, based on compound Bayesian theory, incorporates context by fusing the measurements of all objects under consideration. It is an optimal strategy in terms of achieving minimum set-by-set error probability. The second approach fuses the measurements of an object with explicitly extracted context. Its linear computational complexity makes it more tractable than the first approach, which requires exponential computation. These two techniques are applied to two medical applications: white blood cell image classification and microscopic urinalysis. It is demonstrated that superior classification performances are achieved by using context. In our particular applications, it reduces overall classification error, as well as false positive and false negative diagnosis rates.",
keywords = "Compound Bayesian Theory, Context, Contextual information, Fusion, Image recognition, Pattern recognition",
author = "Xubo Song and Yaser Abu-Mostafa and Joseph Sill and Harvey Kasdan and Misha Pavel",
year = "2002",
month = "12",
doi = "10.1016/S1566-2535(02)00092-1",
language = "English (US)",
volume = "3",
pages = "277--287",
journal = "Information Fusion",
issn = "1566-2535",
publisher = "Elsevier",
number = "4",

}

TY - JOUR

T1 - Robust image recognition by fusion of contextual information

AU - Song, Xubo

AU - Abu-Mostafa, Yaser

AU - Sill, Joseph

AU - Kasdan, Harvey

AU - Pavel, Misha

PY - 2002/12

Y1 - 2002/12

N2 - This paper studies the fusion of contextual information in pattern recognition, with applications to biomedical image identification. In the real world there are cases where the identity of an object is ambiguous if the classification is based only on its own features. It is helpful to reduce the ambiguity by utilizing extra information, referred to as context, provided by accompanying objects. We investigate two techniques that incorporate context. The first approach, based on compound Bayesian theory, incorporates context by fusing the measurements of all objects under consideration. It is an optimal strategy in terms of achieving minimum set-by-set error probability. The second approach fuses the measurements of an object with explicitly extracted context. Its linear computational complexity makes it more tractable than the first approach, which requires exponential computation. These two techniques are applied to two medical applications: white blood cell image classification and microscopic urinalysis. It is demonstrated that superior classification performances are achieved by using context. In our particular applications, it reduces overall classification error, as well as false positive and false negative diagnosis rates.

AB - This paper studies the fusion of contextual information in pattern recognition, with applications to biomedical image identification. In the real world there are cases where the identity of an object is ambiguous if the classification is based only on its own features. It is helpful to reduce the ambiguity by utilizing extra information, referred to as context, provided by accompanying objects. We investigate two techniques that incorporate context. The first approach, based on compound Bayesian theory, incorporates context by fusing the measurements of all objects under consideration. It is an optimal strategy in terms of achieving minimum set-by-set error probability. The second approach fuses the measurements of an object with explicitly extracted context. Its linear computational complexity makes it more tractable than the first approach, which requires exponential computation. These two techniques are applied to two medical applications: white blood cell image classification and microscopic urinalysis. It is demonstrated that superior classification performances are achieved by using context. In our particular applications, it reduces overall classification error, as well as false positive and false negative diagnosis rates.

KW - Compound Bayesian Theory

KW - Context

KW - Contextual information

KW - Fusion

KW - Image recognition

KW - Pattern recognition

UR - http://www.scopus.com/inward/record.url?scp=0036895474&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0036895474&partnerID=8YFLogxK

U2 - 10.1016/S1566-2535(02)00092-1

DO - 10.1016/S1566-2535(02)00092-1

M3 - Article

AN - SCOPUS:0036895474

VL - 3

SP - 277

EP - 287

JO - Information Fusion

JF - Information Fusion

SN - 1566-2535

IS - 4

ER -