基于类内散度的OCR证件图像的倾斜检测算法  

Skew Detection Algorithm for OCR ID Document Image Based on within-group Measure of Dispersion

作　　者：谢国庆[1] 白莹[2] 王智文[3] 

机构地区：[1]福建师范大学软件学院 [2]福建工程学院数理系 [3]广西工学院计算机工程系

出　　处：《广西工学院学报》2009年第2期38-41,46,共5页Journal of Guangxi University of Technology

基　　金：福建省科技厅重点项目(2008H0016)资助;广西教育厅科研项目(200707LX196;200707LX182)资助

摘　　要：针对包含有OCR字符区域的OCR证件图象,提出了一种基于方向投影和类内散度的OCR证件图象的倾斜检测算法.该算法以整数Bresenham算法计算不同扫描角度的方向投影,同时计算各个方向投影的最小总类内散度,然后寻找所有最小总类内散度的最小值,以该最小值对应的扫描角度为最终的检测倾角.实验结果表明,该算法能快速准确地检测出OCR证件图象的倾斜角度.According to the OCR ID Doctment image containing the OCR character region, a new skew detection method based on the directional projection and the within-group measure of dispersion has been presented. Several directional projection are calculated in the method, using Bresenham＇s integer line drawing algorithm. After calculating minimal within-group measure of dispersion of the directional projection and searching minimal value of all within-group measure of dispersion, the corresponding angle of minimal value is the tilted angle of the OCR ID document. Experiment results show that the slanting angle of a OCR ID document image can be detected accurately.

关 键 词：OCR证件 方向投影 类内散度 倾斜检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]