基于条件笔画密度提取的文本定位方法  

Video text location method based on conditional stroke density extraction

作　　者：张建明[1] 王娟[1] 张菊[1] 杜丹[1] 房芳[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013

出　　处：《计算机工程与设计》2011年第10期3446-3449,共4页Computer Engineering and Design

基　　金：江苏省自然科学基金项目(BK2009199)

摘　　要：为了解决视频检索中文本定位精确度不高的问题,提出一种基于条件笔画密度提取的文本定位方法。根据文本笔画的特征对视频图像中的文本进行初步定位,用基于条件密度的方法滤去一部分非文本区域,再利用动态形态学对提取的文本区域精确定位并对其优化得到最终的文本区域。实验结果表明,该方法能滤去大部分非文本信息,得到相对精确的文本区域,并且定位效果较好。In order to improve the curacy of the text location during the video retrieval, a new method of text location approach ispresented based on stroke density extraction. First according to the characters of text strokes, a preliminary location of the text in the video images is obtained. Then the text of video images based on conditional density is used to filter parts of the non-text regions. Finally the dynamic morphology is applied to locate the extracted text, and by optimizing text regions, the last result is gotten. The experimental results show that our method can filter most of non-text regions to obtain relatively accurate text regions and get good localization results in various types of video images.

关 键 词：视频文本定位 笔画提取 条件密度 动态形态学 文本区域 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]