Table structure extraction from form documents based on gradient-wavelet scheme

Dihua Xi, Seong Whan Lee

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

2 Citations (Scopus)


Based on gradient and wavelet analyses, a novel scheme has been developed to extract table structures from skewed form document images. In this scheme, first, a skewed form document image is rotated according to the angle obtained from the gradient algorithm. Then the deskewed image is decomposed into four sub-images by divisible Multiresolution Analysis(MRA) wavelets. Afterwards, the table structure image which represents the geometric structure of the form can be obtained from the sub-images by a modified wavelet reconstruction algorithm. Meanwhile, another document image without table lines can be produced by Minkowski operation and is referred to as a table free image. Experimental results indicate that this new scheme can be applied to process the skewed form document images with promising achievements.

Original languageEnglish
Title of host publicationDocument Analysis Systems
Subtitle of host publicationTheory and Practice - 3rd IAPR Workshop, DAS 1998, Selected Papers
EditorsYasuaki Nakanoc, Seong-Whan Lee
PublisherSpringer Verlag
Number of pages15
ISBN (Print)3540665072, 9783540665076
Publication statusPublished - 1999
Event3rd IAPR Workshop on Document Analysis Systems, DAS 1998 - Nagano, Japan
Duration: 1998 Nov 41998 Nov 6

Publication series

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


Other3rd IAPR Workshop on Document Analysis Systems, DAS 1998

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 1999.

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Table structure extraction from form documents based on gradient-wavelet scheme'. Together they form a unique fingerprint.

Cite this