Features for Neural Net Based Region Identification of Newspaper Documents

  • Authors:
  • Tim Andersen;Wei Zhang

  • Affiliations:
  • -;-

  • Venue:
  • ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Several features for Neural Network based documentregion identification are tested. Specifically, this paperexamines features for non-text region identification. TheNeural Network based region identification algorithm is akey component of a document recognition system thatsegments a document into regions, classifies them intotext, graphic, photo, and other region types, and then usesthis classification to guide the processing and analysis ofthe image. The input data are unusually challenging: lowquality images of newspaper documents obtained frommicrofilmed archives. The results compare favorably withother results reported in the literature.