ABSTRACT
Word segmentation is the preprocessing for word recognition. Word recognition based on characters need that the word is segmented into completely separate characters to be recognized.so word segmentation section must ensure the integrity of the characters to ensure the acuracy of recognition (Chen Tao & Yang Chenhui & Qing Bo 2009). Currently text segmentation methods often used by researchers include vertical projection method and improved one (Jiao Pengpeng & Guo Yizheng 2013, Liu Yangxing 2001, LI Zuo & Wang Shuhua & Cai ShiJie 2001), curve segmenting path method (Liu Yu, & Zhang Yanduo & Lu Tongwei 2011), integrated segmentation method (Meng Qingyuang & Bai Yanping & Hu Hongping 2011), clustering method (WANG J & JEAN J 1994, Wu Rui & Yin Fang & Tang Xianglong, et al 2010) and recognition feedback method (Wang Jiangqing, & Cao Wei 2011). While all these algorithms either can not well process touching and kerned samples or not applicable for printed text recognition because of high complexity, large computational cost and low eciency and accuracy.
