一种新的商务名片的快速二值化算法  

Efficient binary algorithm for business card image under embedded environment

作　　者：刚亚州 黄元元[1] 戴群[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,南京210016

出　　处：《计算机工程与应用》2015年第16期195-198,264,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61100108)

摘　　要：对商务名片进行自动识别,二值化处理是非常关键的一步,其效果将直接影响到后面的版面分析以及字符的分割和识别。而现有的二值化算法时间复杂度高,并且缺乏针对性。提出了一种专门针对商务名片的快速二值化算法,该算法利用颜色模型,通过优化带权误差平方和目标函数找到最优阈值,并给出一个快速迭代算法。经过大量实验证明,相比于传统二值化算法,该算法在嵌入式环境中对商务名片图像做二值化处理,不仅降低了时间复杂度,提高了处理效率,并且对于那些图案设计复杂的名片也起到了很好的预处理的作用,可以去除对于识别结果无用的图案,净化版面,给后续的版面分割等工作打好基础。Binary processing is one of the most important steps in the automatic recognition of business card,its effect will directly affect the subsequent layout analysis,character segmentation and recognition.While the existing binarization algorithm time complexity is very high,and the lack of pertinence.Therefore,a new fast binarization algorithm for a business card is proposed in the paper.This method uses color model and through optimizing an weighted objective function of error sum squares to find the optimal threshold,and what＇s more,a fast iterative algorithm based on it is designed.The experiments show that,compared with the present algorithms,this new algorithm not only reduces the time complexity and improves the processing efficiency in an embedded environment of a business card image binarization processing.And for those complex business card also plays a good pretreatment effect.Thus,can remove the useless pattern recognition results,purification space,and lays a good foundation for the subsequent page segmentation,etc.

关 键 词：嵌入式 名片识别 二值化处理 目标函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]