基于Gabor滤波与边缘特征的场景文字检测  

SCENE TEXT DETECTION BASED ON GABOR FILTERING AND EDGE FEATURES

作　　者：邓勇[1] 武小红[1] 雍宝虎[1] 

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013

出　　处：《计算机应用与软件》2012年第12期286-290,共5页Computer Applications and Software

摘　　要：为了解决场景文字检测受到背景复杂、光照不均等环境因素及文字本身属性变化的影响,提出一种基于Gabor滤波与边缘特征的场景文字检测方法。有效地融合了图像的纹理特征与边缘信息来得到文本候选区;经过形态学处理与投影运算对文字区域进行定位;最后通过一种特殊的二值化处理实现文字的提取。实验表明,该方法能检测各种场景下的文字。In order to solve the problem of scene text detection that it is impacted by the environmental factors of complex background and uneven illumination as well as the variations of attribute of text＇ s own, we present a text detection approach which is based on Gabor filtering and edge features. The new method attains candidate text regions by effectively integrating the texture feature and edge information of the im- age ; locates the text regions through morphological processing and projection operation; and finally achieves text extraction using a special binarisation processing. Experiments show that this method can well detect the texts in various scenes.

关 键 词：场景文字 GABOR滤波 边缘特征 形态学处理 投影运算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]