Improving Thai Optical Character Recognition Using Circular-Scan Histogram

Natsuda Kaothanthong, Thanaruk Theeramunkong, Jinhee Chun

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

1 Citation (Scopus)

Abstract

While most previous works focus on invariant features that capture salient for higher recognition accuracy, they face an issue of misclassification among similar characters. This paper proposes a feature called circular-scan histogram that enables us to capture small salient parts of the Thai characters. With scanning distances, the distance from the edge to the first pixel on the character's boundary, a circular-scan histogram is constructed by rotating the characters during scanning, and counting frequency of each distance bin. By experiments, using approximately 60,000 single-character images of forty four Thai consonant characters with balance distribution, our proposed method can classify similar characters with accuracy of 95.72%. As baselines, we compare our method with Shape Context and Histogram of Oriented Gradient.

Original languageEnglish
Title of host publicationProceedings - 14th IAPR International Conference on Document Analysis and Recognition, ICDAR 2017
PublisherIEEE Computer Society
Pages567-572
Number of pages6
ISBN (Electronic)9781538635865
DOIs
Publication statusPublished - 2017 Jul 2
Externally publishedYes
Event14th IAPR International Conference on Document Analysis and Recognition, ICDAR 2017 - Kyoto, Japan
Duration: 2017 Nov 92017 Nov 15

Publication series

NameProceedings of the International Conference on Document Analysis and Recognition, ICDAR
Volume1
ISSN (Print)1520-5363

Other

Other14th IAPR International Conference on Document Analysis and Recognition, ICDAR 2017
CountryJapan
CityKyoto
Period17/11/917/11/15

Keywords

  • Circular-Scan Distance
  • Optical Character Recognition
  • Thai OCR

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Fingerprint Dive into the research topics of 'Improving Thai Optical Character Recognition Using Circular-Scan Histogram'. Together they form a unique fingerprint.

Cite this