Recognition of Handwritten Numerals Based on Structural Features
Luzheng Bi · Jisuanji gongcheng · 2002
In this paper, a new method based on structural features for the recognition of handwritten numerals is proposed. Firstly, an algorithm for thinning with 14 pieces of template is presented. Secondly, point, line and circle features cell are extracted and combined. And finally, a two-stage recognition method is put forward, in which primitive classification and template match based on distance measures are introduced.