A study on the document zone content classification problem

Yalin Wang, Ihsin T. Phillips, Robert M. Haralick

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

A document can be divided into zones on the basis of its content. For example, a zone can be either text or non-text. Given the segmented document zones, correctly determining the zone content type is very important for the subsequent processes within any document image understanding system. This paper describes an algorithm for the determination of zone type of a given zone within an input document image. In our zone classification algorithm, zones are represented as feature vectors. Each feature vector consists of a set of 25 measurements of pre-defined properties. A probabilistic model, decision tree, is used to classify each zone on the basis of its feature vector. Two methods are used to optimize the decision tree classifier to eliminate the data over-fitting problem. To enrich our probabilistic model, we incorporate context constraints for certain zones within their neighboring zones.We also model zone class context constraints as a Hidden Markov Model and usedViterbi algorithm to obtain optimal classification results. The training, pruning and testing data set for the algorithm include 1, 600 images drawn from theUWCDROM-III document image database.With a total of 24, 177 zones within the data set, the cross-validation methodwas used in the performance evaluation of the classifier. The classifier is able to classify each given scientific and technical document zone into one of the nine classes, 2 text classes (of font size 4-18pt and font size 19-32 pt), math, table, halftone, map/drawing, ruling, logo, and others. A zone content classification performance evaluation protocol is proposed. Using this protocol, our algorithm accuracy is 98.45% with a mean false alarm rate of 0.50%.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages212-223
Number of pages12
Volume2423
ISBN (Print)3540440682, 9783540440680
StatePublished - 2002
Externally publishedYes
Event5th International Workshop on Document Analysis Systems, DAS 2002 - Princeton, United States
Duration: Aug 19 2002Aug 21 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2423
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other5th International Workshop on Document Analysis Systems, DAS 2002
CountryUnited States
CityPrinceton
Period8/19/028/21/02

Fingerprint

Classification Problems
Classifiers
Decision trees
Drawing (graphics)
Image understanding
Hidden Markov models
Feature Vector
Classifier
Decision tree
Probabilistic Model
Performance Evaluation
Testing
Classify
Image Understanding
False Alarm Rate
Overfitting
Image Database
Classification Algorithm
Pruning
Cross-validation

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Wang, Y., Phillips, I. T., & Haralick, R. M. (2002). A study on the document zone content classification problem. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 2423, pp. 212-223). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2423). Springer Verlag.

A study on the document zone content classification problem. / Wang, Yalin; Phillips, Ihsin T.; Haralick, Robert M.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2423 Springer Verlag, 2002. p. 212-223 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2423).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Wang, Y, Phillips, IT & Haralick, RM 2002, A study on the document zone content classification problem. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 2423, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 2423, Springer Verlag, pp. 212-223, 5th International Workshop on Document Analysis Systems, DAS 2002, Princeton, United States, 8/19/02.
Wang Y, Phillips IT, Haralick RM. A study on the document zone content classification problem. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2423. Springer Verlag. 2002. p. 212-223. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Wang, Yalin ; Phillips, Ihsin T. ; Haralick, Robert M. / A study on the document zone content classification problem. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2423 Springer Verlag, 2002. pp. 212-223 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{c4afdc9762644204a999e62bfb781ec3,
title = "A study on the document zone content classification problem",
abstract = "A document can be divided into zones on the basis of its content. For example, a zone can be either text or non-text. Given the segmented document zones, correctly determining the zone content type is very important for the subsequent processes within any document image understanding system. This paper describes an algorithm for the determination of zone type of a given zone within an input document image. In our zone classification algorithm, zones are represented as feature vectors. Each feature vector consists of a set of 25 measurements of pre-defined properties. A probabilistic model, decision tree, is used to classify each zone on the basis of its feature vector. Two methods are used to optimize the decision tree classifier to eliminate the data over-fitting problem. To enrich our probabilistic model, we incorporate context constraints for certain zones within their neighboring zones.We also model zone class context constraints as a Hidden Markov Model and usedViterbi algorithm to obtain optimal classification results. The training, pruning and testing data set for the algorithm include 1, 600 images drawn from theUWCDROM-III document image database.With a total of 24, 177 zones within the data set, the cross-validation methodwas used in the performance evaluation of the classifier. The classifier is able to classify each given scientific and technical document zone into one of the nine classes, 2 text classes (of font size 4-18pt and font size 19-32 pt), math, table, halftone, map/drawing, ruling, logo, and others. A zone content classification performance evaluation protocol is proposed. Using this protocol, our algorithm accuracy is 98.45{\%} with a mean false alarm rate of 0.50{\%}.",
author = "Yalin Wang and Phillips, {Ihsin T.} and Haralick, {Robert M.}",
year = "2002",
language = "English (US)",
isbn = "3540440682",
volume = "2423",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "212--223",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - A study on the document zone content classification problem

AU - Wang, Yalin

AU - Phillips, Ihsin T.

AU - Haralick, Robert M.

PY - 2002

Y1 - 2002

N2 - A document can be divided into zones on the basis of its content. For example, a zone can be either text or non-text. Given the segmented document zones, correctly determining the zone content type is very important for the subsequent processes within any document image understanding system. This paper describes an algorithm for the determination of zone type of a given zone within an input document image. In our zone classification algorithm, zones are represented as feature vectors. Each feature vector consists of a set of 25 measurements of pre-defined properties. A probabilistic model, decision tree, is used to classify each zone on the basis of its feature vector. Two methods are used to optimize the decision tree classifier to eliminate the data over-fitting problem. To enrich our probabilistic model, we incorporate context constraints for certain zones within their neighboring zones.We also model zone class context constraints as a Hidden Markov Model and usedViterbi algorithm to obtain optimal classification results. The training, pruning and testing data set for the algorithm include 1, 600 images drawn from theUWCDROM-III document image database.With a total of 24, 177 zones within the data set, the cross-validation methodwas used in the performance evaluation of the classifier. The classifier is able to classify each given scientific and technical document zone into one of the nine classes, 2 text classes (of font size 4-18pt and font size 19-32 pt), math, table, halftone, map/drawing, ruling, logo, and others. A zone content classification performance evaluation protocol is proposed. Using this protocol, our algorithm accuracy is 98.45% with a mean false alarm rate of 0.50%.

AB - A document can be divided into zones on the basis of its content. For example, a zone can be either text or non-text. Given the segmented document zones, correctly determining the zone content type is very important for the subsequent processes within any document image understanding system. This paper describes an algorithm for the determination of zone type of a given zone within an input document image. In our zone classification algorithm, zones are represented as feature vectors. Each feature vector consists of a set of 25 measurements of pre-defined properties. A probabilistic model, decision tree, is used to classify each zone on the basis of its feature vector. Two methods are used to optimize the decision tree classifier to eliminate the data over-fitting problem. To enrich our probabilistic model, we incorporate context constraints for certain zones within their neighboring zones.We also model zone class context constraints as a Hidden Markov Model and usedViterbi algorithm to obtain optimal classification results. The training, pruning and testing data set for the algorithm include 1, 600 images drawn from theUWCDROM-III document image database.With a total of 24, 177 zones within the data set, the cross-validation methodwas used in the performance evaluation of the classifier. The classifier is able to classify each given scientific and technical document zone into one of the nine classes, 2 text classes (of font size 4-18pt and font size 19-32 pt), math, table, halftone, map/drawing, ruling, logo, and others. A zone content classification performance evaluation protocol is proposed. Using this protocol, our algorithm accuracy is 98.45% with a mean false alarm rate of 0.50%.

UR - http://www.scopus.com/inward/record.url?scp=84947757232&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84947757232&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84947757232

SN - 3540440682

SN - 9783540440680

VL - 2423

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 212

EP - 223

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -