计算机应用 ›› 2010, Vol. 30 ›› Issue (12): 3325-3326.

• 虚拟现实与模式识别 • 上一篇    下一篇

复杂背景下的号码定位与分割

邱炜1,陈斌2   

  1. 1. 成都计算所
    2. 中国科学院成都计算机应用研究所
  • 收稿日期:2010-05-05 修回日期:2010-06-03 发布日期:2010-12-22 出版日期:2010-12-01
  • 通讯作者: 邱炜

Number location and character segmentation under complex background

  • Received:2010-05-05 Revised:2010-06-03 Online:2010-12-22 Published:2010-12-01

摘要: 提出一种综合边缘检测、号码分布特点的号码定位方法和基于投影及最大连通区域的字符分割方法。采用改进的数学形态学方法提取灰度图像边缘,用最小二乘法对边缘进行直线拟合得到图像的倾斜角,进而对倾斜的图像进行校正,再利用号码的位置信息来定位号码区域。采用最大连通区域算法过滤块状噪声,通过投影和字符的宽度特点确定字符的分割结果,有效地解决了由复杂背景、油墨的深浅、污迹、磨损带来的干扰。

关键词: 边缘检测, 倾斜校正, 号码定位, 字符分割, 最大连通区域

Abstract: In this paper, an edge detection combined with number distribution algorithm was proposed to locate the number area, and a projection-maximal connected region algorithm was proposed to segment characters. An improved mathematical morphology was used to extract the edge of gray image; the max connected region algorithm was used to figure out the skew angle for image skew correction. After that, number area could be located with the help of number location information. The maximal connected region algorithm was adopted to filter block noise, and projection feature and width of character were used to segment characters. This method effectively overcomes the disturbance due to complex background, the depth of ink, stains and wear.

Key words: edge detection, skew correction, number location, character segmentation, maximal connected region