Multiresolution image classification by hierarchical modeling with two-dimensional hidden Markov models

Jia Li, Robert M. Gray, Richard A. Olshen

Research output: Contribution to journalArticle

97 Citations (Scopus)

Abstract

This paper treats a multiresolution hidden Markov model for classifying images. Each image is represented by feature vectors at several resolutions, which are statistically dependent as modeled by the underlying state process, a multiscale Markov mesh. Unknowns in the model are estimated by maximum likelihood, in particular by employing the expectation-maximization algorithm. An image is classified by finding the optimal set of states with maximum a posteriori probability. States are then mapped into classes. The multiresolution model enables multiscale information about context to be incorporated into classification. Suboptimal algorithms based on the model provide progressive classification that is much faster than the algorithm based on single-resolution hidden Markov models.

Original languageEnglish (US)
Pages (from-to)1826-1841
Number of pages16
JournalIEEE Transactions on Information Theory
Volume46
Issue number5
DOIs
StatePublished - Aug 2000

Fingerprint

Image classification
Hidden Markov models
Maximum likelihood

All Science Journal Classification (ASJC) codes

  • Information Systems
  • Computer Science Applications
  • Library and Information Sciences

Cite this

@article{57a47fc225374abd8aeb2d7baa7a99c1,
title = "Multiresolution image classification by hierarchical modeling with two-dimensional hidden Markov models",
abstract = "This paper treats a multiresolution hidden Markov model for classifying images. Each image is represented by feature vectors at several resolutions, which are statistically dependent as modeled by the underlying state process, a multiscale Markov mesh. Unknowns in the model are estimated by maximum likelihood, in particular by employing the expectation-maximization algorithm. An image is classified by finding the optimal set of states with maximum a posteriori probability. States are then mapped into classes. The multiresolution model enables multiscale information about context to be incorporated into classification. Suboptimal algorithms based on the model provide progressive classification that is much faster than the algorithm based on single-resolution hidden Markov models.",
author = "Jia Li and Gray, {Robert M.} and Olshen, {Richard A.}",
year = "2000",
month = "8",
doi = "10.1109/18.857794",
language = "English (US)",
volume = "46",
pages = "1826--1841",
journal = "IEEE Transactions on Information Theory",
issn = "0018-9448",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "5",

}

Multiresolution image classification by hierarchical modeling with two-dimensional hidden Markov models. / Li, Jia; Gray, Robert M.; Olshen, Richard A.

In: IEEE Transactions on Information Theory, Vol. 46, No. 5, 08.2000, p. 1826-1841.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Multiresolution image classification by hierarchical modeling with two-dimensional hidden Markov models

AU - Li, Jia

AU - Gray, Robert M.

AU - Olshen, Richard A.

PY - 2000/8

Y1 - 2000/8

N2 - This paper treats a multiresolution hidden Markov model for classifying images. Each image is represented by feature vectors at several resolutions, which are statistically dependent as modeled by the underlying state process, a multiscale Markov mesh. Unknowns in the model are estimated by maximum likelihood, in particular by employing the expectation-maximization algorithm. An image is classified by finding the optimal set of states with maximum a posteriori probability. States are then mapped into classes. The multiresolution model enables multiscale information about context to be incorporated into classification. Suboptimal algorithms based on the model provide progressive classification that is much faster than the algorithm based on single-resolution hidden Markov models.

AB - This paper treats a multiresolution hidden Markov model for classifying images. Each image is represented by feature vectors at several resolutions, which are statistically dependent as modeled by the underlying state process, a multiscale Markov mesh. Unknowns in the model are estimated by maximum likelihood, in particular by employing the expectation-maximization algorithm. An image is classified by finding the optimal set of states with maximum a posteriori probability. States are then mapped into classes. The multiresolution model enables multiscale information about context to be incorporated into classification. Suboptimal algorithms based on the model provide progressive classification that is much faster than the algorithm based on single-resolution hidden Markov models.

UR - http://www.scopus.com/inward/record.url?scp=0034245909&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0034245909&partnerID=8YFLogxK

U2 - 10.1109/18.857794

DO - 10.1109/18.857794

M3 - Article

AN - SCOPUS:0034245909

VL - 46

SP - 1826

EP - 1841

JO - IEEE Transactions on Information Theory

JF - IEEE Transactions on Information Theory

SN - 0018-9448

IS - 5

ER -