基于mean shift算法的文字分割及应用  

Algorithm for Text Segmentation and Its Application Based on Mean Shift

作　　者：杨富元[1] 周晓东[1] 

机构地区：[1]长春工业大学机械工程学院,长春130012

出　　处：《机械工程师》2014年第1期42-44,共3页Mechanical Engineer

基　　金：吉林省科技厅重点项目(20120351)

摘　　要：传统的文字识别中,文字分割的计算量大,使文字识别速度慢。mean shift算法利用直方图描述和寻找分割区域,算法简单且具有很好的鲁棒性。改进后的mean shift算法,利用字符串间的距离信息及完整字符串经过屏幕所需的时间信息进行预测位置,减少了迭代次数,因此运算速度快,可实时进行跟踪字符串和连续分割。实验结果表明:该算法可以对连续的字符串进行稳定、实时的文字分割。Text segmentation computation is large in traditional text recognition method, which causes slow character reeognition speed.Mean shift algorithm is simple and has good robustness using the histogram description and seek segmented regions. Improved mean shift algorithm can predict position using the distance between the strings and the time of the full strings pass the sreen .The method reduces the number of iterations, so it can track strings in real time and split strings continuously. Exprimental results show that the algorithm can split continuous strings stably in real time.

关 键 词：均值漂移 文字分割 文字识别 相似性度量 直方图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]