ApOFIS: an A priori optical font identification system

  • Abdelwahab Zramdini
  • Rolf Ingold
Pattern Recognition and Document Processing
Part of the Lecture Notes in Computer Science book series (LNCS, volume 974)


The detection of the font style, point size, etc. of a text is an obvious way to improve the capabilities of text recognition algorithms. The ApOFIS system has been designed in order to satisfy such a requirement. It adopts an a priori font identification approach where the recognition of a text font is done without considering the characters that appear in the text. In ApOFIS, a font is characterized especially by its family, weight, slope and size. Features used in the system represent global aspects of text line images. They have been extracted essentially from projection profiles and from connected components bounding boxes. Statistical tests have revealed that these features follow approximately normal laws so that parameter estimation is used in learning.

A multivariate Bayesian classifier, based on these features, has been designed for font recognition and applied on a base of 240 font models created from a training set of texts written with these fonts. On text lines having the same length as those used for learning, the system allows to discriminate fonts with an average accuracy of 96.5% for top choice and 98.3% within the two top choices.


Document Analysis a priori Font Recognition Bayesian Classifier 


  1. 1.
    T. Bayer, J. Hull, and G. Nagy, ‘Character Recognition: SSPR'90 working group report', in Structured Document Image Analysis, eds., H.S. Baird, H. Bunke, and K. Yamamoto, 567–567, Springer Verlag, (1992).Google Scholar
  2. 2.
    A. Zramdini and R. Ingold, ‘Optical font recognition from projection profiles', in RIDT'94: Third International Conference on Raster Imaging and Digital Typography, pp. 249–260, Darmstadt, Germany, (4 1994).Google Scholar
  3. 3.
    R. A. Morris, ‘Classification of digital typefaces using spectral signatures', Pattern Recognition, 25(8), 869–876, (1988).Google Scholar
  4. 4.
    Tao Hu, New Methods for Robust and Efficient Recognition of the Logical Structures in Documents, Ph.D. dissertation, University of Fribourg, 1994.Google Scholar
  5. 5.
    A. Zramdini and R. Ingold, ‘A priori font recognition using a Bayesain classifier', Technical report, IIUF, University of Fribourg, Switzerland, (2 1994).Google Scholar
  6. 6.
    A. Zramdini and R. Ingold, ‘A Study of Document Image Degradation Effects on Font Recognition', to be published in ICDAR'95: Third International Conference on Document Analysis and Recognition, Montreal, Canada, (8 1995).Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1995

Authors and Affiliations

  • Abdelwahab Zramdini
    • 1
  • Rolf Ingold
    • 1
  1. 1.Institute of InformaticsUniversity of FribourgFribourgSwitzerland

Personalised recommendations