A Database of Handwritten Characters for Writer Recognition Study and Software for Display and Analysis

Abstract

  A new database and software for the study of writer recognition was created. The database of handwritten characters is based on 100 character types written 50 times by 100 writers. The software, called Characom Imager 100 (Character data Computer-processing Imager 100), is capable of displaying the collected characters on the computer screen and comparing character forms while conducting basic image processing.
  Previously, such an enormous database of handwritten characters that could be used for writer recognition study did not exist. This common, large-scale database will facilitate comparison and evaluation based on data from other studies and will contribute to the improvement of writer recognition technique and reliability.
  Characom Imager 100 is capable of quickly searching out a character from a vast database and displaying it on the computer screen under prescribed conditions, with the center of the character pattern placed on the center line of the display screen. The software carries out a comparative analysis of characters while conducting seven types of image processing: sampling, border following, distance transformation, labeling, thinning, projection profiling and feature extraction. This software adds new dimensions to the comparison of character patterns. The characteristics of characters obtained through image processing can be used to analyze character forms and can be translated into graphs or figures for analysis. Application of this software could produce new character-identification methods not only in terms of writer recognition, but also in terms of handwriting identification.