有效的场景文本提取算法  

Effective algorithm of text extraction in natural scenes

作　　者：何兴恒[1] 胡德婷[1] 

机构地区：[1]中国地质大学(武汉)计算机学院,湖北武汉430074

出　　处：《计算机工程与设计》2008年第10期2598-2599,2603,共3页Computer Engineering and Design

摘　　要：自然场景图像中的文本提供了重要的语意信息,它是图像内容的重要来源。针对当前的求解算法普遍存在提取文本精确度不高等缺点,提出了一种文本定位准确的文本提取算法。先将原始图片进行金字塔分解,然后进行彩色图像边缘提取和二值化,再形态学文本定位,最后文本区域字符提取。对ICDAR数据库图片的测试结果表明,该方法对文字颜色、大小字体以及排列方向具有较强的鲁棒性,同时也具有较高的精确度和提取率。The text of scene images can provide an important index of semantic information, It is the key source of the image content. In order to solve the problems of methods which are used nowadays, such as low precision, an efficient way of text extraction from scene images is presented. First, pyramid decomposition is used through the original image. Then, the color image edge detection and binarization is used. Through the mathematical morphology method, the text of the color image is extracted by the restrictions of text regions at last. From a large number of images testing in ICDAR＇s database, this algorithm shows its robustness and accuracy in text color, font size and configuration from complex background.

关 键 词：文本提取 金字塔分解 边缘检测 二值化 数学形态学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]