Most of Hangul recognition systems understand only single font documents. The purpose of this thesis is to develop a method that removes the font restrictions as much as possible while maintaining high recognition rate. A Hangul character consists of a phonetic alphabet containing consonants and vowels. The fact that the number of Hangul alphabet is remarkably small in compared with that of Hangul character is the motive that we adopt an algorithm for the recognition of Hangul with a Hangul separated into alphabets. Alphabet separation is a key stage for the proposed Hangul recognition algorithm.
With piecewise linear approximation of the contour of a Hangul image, information on the existence of long horizontal (or vertical) stroke of vowel is used to separate the alphabets of Hangul. Features are extracted from the contour of the separated alphabets, and the fonts as well as the alphabets are recognized by matching these features with the alphabet models which are constructed in a training stage.
In a test case with the most frequently used 522 printed Hanguls in Myungjo, Sinmun, Gothic, Dinaru, Gungseo and Gongjak fonts, it achieved 95% of alphabet separation rate, 96% of character recognition rate and 91% of font recognition rate. While it still has a room for improvement, the promise of our approach was demonstrated.
