一种多字体印刷藏文字符的归一化方法  

A Normalization Method of Multi-font Printed Tibetan Characters

作　　者：王华[1] 丁晓青[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《计算机应用研究》2004年第6期41-43,共3页Application Research of Computers

基　　金：国家"86 3"高技术计划资助项目 (2 0 0 1AA1 1 4 0 81 ) ;国家自然科学基金资助项目 (6 9972 0 2 4 )

摘　　要：消除输入字符在位置和大小上的差异的归一化操作是字符识别系统中一个重要环节。在详细分析藏文字符字形特征的基础上 ,提出了一种多字体印刷藏文字符归一化方法 :综合运用字符重心和外边框信息实现位置归一化 ,然后采用三次B样条函数将字符归一化到 48× 96的目标点阵。In an OCR (Optical Character Recognition) system,character normalization is a crucial step to eliminate variations in character size or position.In this paper,based on the detailed analysis of the characteristics of shape and stroke distribution of multi-font printed tibetan characters,a new normalization algorithm for tibetan OCR is proposed.Firstly character position is normalized combing profile information with the centroid of input character images.Then the 48×96 block is introduced to perform the size normalization by cubic B-spline.The effectiveness of proposed algorithm is demonstrated by experimental results.

关 键 词：藏文字符识别 归一化 基线 三次B样条 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]