提高OCR识别率的研究  

A Research on How to Improve the Recognition Rate of OCR

作　　者：白廷柱[1] 

机构地区：[1]北京理工大学工程光学系

出　　处：《北京理工大学学报》1994年第S1期96-100,共5页Transactions of Beijing Institute of Technology

摘　　要：根据ＯＣＲ（ＯｐｔｉｃａｌＣｈａｒａｃｔｅｒＲｅａｄｅｒ）读入图象象素灰度的统计分布关系，利用象素灰度分布的类间方差，找出使象素灰度分布的类间方差为最大的象素灰度级，并以此作为图象的二值化阈值，对由ＯＣＲ读入的帐票图象等进行二值化处理，以解决人工设定阈值法所带来的文字识别率下降的问题．According to the probability and statistics law of the image element,by calculating the between-class variance, in this paper,the image element gray level with the greatest between-class variance has been found out, and being the binarization threshold in the binarizing solution to the bills read by OCR, it fairly solves the recognition dropping problem introduced by the traditional threshold method binzrization.

关 键 词：图象分析 灰度/二值化阈值 类间方差 

分 类 号：O235[理学—运筹学与控制论]