›› 2016, Vol. 29 ›› Issue (10): 39-.

• 论文 • 上一篇    下一篇

基于文本行基线的倾斜角检测算法

巨志勇,何晓蕾,王超男   

  1. (上海理工大学 上海现代光学系统重点实验室,上海 200093)
  • 出版日期:2016-10-15 发布日期:2016-11-14
  • 作者简介:巨志勇(1975-),男,博士,讲师。研究方向:图像处理和模式识别。何晓蕾(1991-),女,硕士研究生。研究方向:图像处理和模式识别。王超男(1992-),女,硕士研究生。研究方向:图像处理和模式识别。
  • 基金资助:

    国家自然科学基金资助项目(81101116)

A Skew Detection Algorithm Based on the Baseline of Document

JU Zhiyong, HE Xiaolei, WANG Chaonan   

  1. (Shanghai Key Laboratory of Modern Optical System, University of Shanghai for Science & Technology, Shanghai 200093, China)
  • Online:2016-10-15 Published:2016-11-14

摘要:

针对文本图像倾斜检测问题,提出了一种新的基于文本行基线的倾斜角检测算法。该算法用边界标记自动机对一组同行的字符进行轮廓(外边界)跟踪,并标记出字符的最小外接矩形(MER)和字符的边框。在此基础上通过相邻字符间的行高差和字符区域的面积来剔除字符的冗余部分,最后用剩余部分的字符边框底边中点来拟合一条直线,即行文本的基线,并确定文本的倾斜角度。实验结果表明,该方法有效,同时倾斜角检测的精确性得到了优化。

关键词: 文本图像, 倾斜检测, 边界标定, 直线拟合

Abstract:

For the problem, this paper proposed a new skew detection algorithm based on the baseline of document for detecting the angle of the document image. The algorithm obtains the minimum external rectangles and the frames of the characters by the regionlabelingautomata, and eliminates the redundancy of characters based on the areas of the parts and the line height difference of the adjacent characters. A straight line is fitted as the baseline based on the frame of rest characters, and the skew angle is evaluated by the moment calculation. The experimental results show that this algorithm is effective and can improve the accuracy.

Key words: document image, skew detection, boundary demarcation, linear fitting

中图分类号: 

  • TP391.41