A hierarchical visual saliency model for character detection in natural scenes

Renwu Gao, Faisal Shafait, Seiichi Uchida, Yaokai Feng

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

4 Citations (Scopus)

Abstract

Visual saliency models have been introduced to the field of character recognition for detecting characters in natural scenes. Researchers believe that characters have different visual properties from their non-character neighbors, which make them salient. With this assumption, characters should response well to computational models of visual saliency. However in some situations, characters belonging to scene text mignt not be as salient as one might expect. For instance, a signboard is usually very salient but the characters on the signboard might not necessarily be so salient globally. In order to analyze this hypothesis in more depth, we first give a view of how much these background regions, such as sign boards, affect the task of saliency-based character detection in natural scenes. Then we propose a hierarchical-saliency method for detecting characters in natural scenes. Experiments on a dataset with over 3,000 images containing scene text show that when using saliency alone for scene text detection, our proposed hierarchical method is able to capture a larger percentage of text pixels as compared to the conventional single-pass algorithm.

Original languageEnglish
Title of host publicationCamera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers
PublisherSpringer Verlag
Pages18-29
Number of pages12
ISBN (Print)9783319051666
DOIs
Publication statusPublished - Jan 1 2014
Event5th International Workshop on Camera-Based Document Analysis and Recognition, CBDAR 2013 - Washington, DC, United States
Duration: Aug 23 2013Aug 23 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8357 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other5th International Workshop on Camera-Based Document Analysis and Recognition, CBDAR 2013
CountryUnited States
CityWashington, DC
Period8/23/138/23/13

Fingerprint

Saliency
Character recognition
Pixels
Model
Experiments
Character Recognition
Vision
Character
Computational Model
Percentage
Pixel
Text
Experiment

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Gao, R., Shafait, F., Uchida, S., & Feng, Y. (2014). A hierarchical visual saliency model for character detection in natural scenes. In Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers (pp. 18-29). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8357 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-05167-3_2

A hierarchical visual saliency model for character detection in natural scenes. / Gao, Renwu; Shafait, Faisal; Uchida, Seiichi; Feng, Yaokai.

Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers. Springer Verlag, 2014. p. 18-29 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8357 LNCS).

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

Gao, R, Shafait, F, Uchida, S & Feng, Y 2014, A hierarchical visual saliency model for character detection in natural scenes. in Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8357 LNCS, Springer Verlag, pp. 18-29, 5th International Workshop on Camera-Based Document Analysis and Recognition, CBDAR 2013, Washington, DC, United States, 8/23/13. https://doi.org/10.1007/978-3-319-05167-3_2
Gao R, Shafait F, Uchida S, Feng Y. A hierarchical visual saliency model for character detection in natural scenes. In Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers. Springer Verlag. 2014. p. 18-29. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-05167-3_2
Gao, Renwu ; Shafait, Faisal ; Uchida, Seiichi ; Feng, Yaokai. / A hierarchical visual saliency model for character detection in natural scenes. Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers. Springer Verlag, 2014. pp. 18-29 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{07e4fdc601d345ee982e971209d84247,
title = "A hierarchical visual saliency model for character detection in natural scenes",
abstract = "Visual saliency models have been introduced to the field of character recognition for detecting characters in natural scenes. Researchers believe that characters have different visual properties from their non-character neighbors, which make them salient. With this assumption, characters should response well to computational models of visual saliency. However in some situations, characters belonging to scene text mignt not be as salient as one might expect. For instance, a signboard is usually very salient but the characters on the signboard might not necessarily be so salient globally. In order to analyze this hypothesis in more depth, we first give a view of how much these background regions, such as sign boards, affect the task of saliency-based character detection in natural scenes. Then we propose a hierarchical-saliency method for detecting characters in natural scenes. Experiments on a dataset with over 3,000 images containing scene text show that when using saliency alone for scene text detection, our proposed hierarchical method is able to capture a larger percentage of text pixels as compared to the conventional single-pass algorithm.",
author = "Renwu Gao and Faisal Shafait and Seiichi Uchida and Yaokai Feng",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/978-3-319-05167-3_2",
language = "English",
isbn = "9783319051666",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "18--29",
booktitle = "Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers",
address = "Germany",

}

TY - GEN

T1 - A hierarchical visual saliency model for character detection in natural scenes

AU - Gao, Renwu

AU - Shafait, Faisal

AU - Uchida, Seiichi

AU - Feng, Yaokai

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Visual saliency models have been introduced to the field of character recognition for detecting characters in natural scenes. Researchers believe that characters have different visual properties from their non-character neighbors, which make them salient. With this assumption, characters should response well to computational models of visual saliency. However in some situations, characters belonging to scene text mignt not be as salient as one might expect. For instance, a signboard is usually very salient but the characters on the signboard might not necessarily be so salient globally. In order to analyze this hypothesis in more depth, we first give a view of how much these background regions, such as sign boards, affect the task of saliency-based character detection in natural scenes. Then we propose a hierarchical-saliency method for detecting characters in natural scenes. Experiments on a dataset with over 3,000 images containing scene text show that when using saliency alone for scene text detection, our proposed hierarchical method is able to capture a larger percentage of text pixels as compared to the conventional single-pass algorithm.

AB - Visual saliency models have been introduced to the field of character recognition for detecting characters in natural scenes. Researchers believe that characters have different visual properties from their non-character neighbors, which make them salient. With this assumption, characters should response well to computational models of visual saliency. However in some situations, characters belonging to scene text mignt not be as salient as one might expect. For instance, a signboard is usually very salient but the characters on the signboard might not necessarily be so salient globally. In order to analyze this hypothesis in more depth, we first give a view of how much these background regions, such as sign boards, affect the task of saliency-based character detection in natural scenes. Then we propose a hierarchical-saliency method for detecting characters in natural scenes. Experiments on a dataset with over 3,000 images containing scene text show that when using saliency alone for scene text detection, our proposed hierarchical method is able to capture a larger percentage of text pixels as compared to the conventional single-pass algorithm.

UR - http://www.scopus.com/inward/record.url?scp=84958534245&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84958534245&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-05167-3_2

DO - 10.1007/978-3-319-05167-3_2

M3 - Conference contribution

AN - SCOPUS:84958534245

SN - 9783319051666

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

SP - 18

EP - 29

BT - Camera-Based Document Analysis and Recognition - 5th International Workshop, CBDAR 2013, Revised Selected Papers

PB - Springer Verlag

ER -