Text extraction from mail images with complex background

  • Qingqing Wang*
  • , Xiao Tu
  • , Shujing Lu
  • , Yue Lu
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

A novel method is proposed for text extraction from mail images with complex background. Firstly, wavelet transform and Laplacian operator are applied to generate the features of regions which are obtained by dividing input image with sliding window. Then, support vector machine (SVM) is utilized to classify these regions into texts and non-texts according to the features. Bootstrap strategy is used to build the training database. Finally, connected components analysis (CCA) is employed to merge text regions into text candidates which can be processed by following steps to get the delivery address. Experimental results involving 534 mail images show the effectiveness and robustness of the proposed method, and comparison results with other methods demonstrate the advantages of the selected features.

Original languageEnglish
Title of host publicationDigital TV and Wireless Multimedia Communication - 14th International Forum, IFTC 2017, Revised Selected Papers
EditorsXiaokang Yang, Guangtao Zhai, Jun Zhou
PublisherSpringer Verlag
Pages3-11
Number of pages9
ISBN (Print)9789811081071
DOIs
StatePublished - 2018
Event14th International Forum of Digital TV and Wireless Multimedia Communication, IFTC 2017 - Shanghai, China
Duration: 8 Nov 20179 Nov 2017

Publication series

NameCommunications in Computer and Information Science
Volume815
ISSN (Print)1865-0929

Conference

Conference14th International Forum of Digital TV and Wireless Multimedia Communication, IFTC 2017
Country/TerritoryChina
CityShanghai
Period8/11/179/11/17

Keywords

  • Mail images
  • SVM
  • Text extraction
  • Wavelet transform

Fingerprint

Dive into the research topics of 'Text extraction from mail images with complex background'. Together they form a unique fingerprint.

Cite this