欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 视觉与图像 • 上一篇    下一篇

一种基于页眉线的扭曲文档图像快速校正方法

  

  • 出版日期:2016-02-26 发布日期:2016-02-26

A Correcting Method Based on Header and Footer Line for Warped Documnet Images

  • Online:2016-02-26 Published:2016-02-26

PDF (PC)

332

被引次数

2

摘要： ：在对文档图像进行光学字符识别时，由于书籍扭曲的存在，识别率会降低。对于
含有页眉页脚线的扭曲文档图像，提出一种快速校正方法。首先分别检测并定位图像中的页眉
线，保存页眉线的坐标信息。根据等比算法计算页眉线上各点在校正时所需向上或向下移动的
距离，然后以此距离为参数扫描图像，计算页眉页脚线之间的各个目标像素校正所需移动的距
离，同时进行像素点的移动重构图像，最终得到校正的图像。实验结果表明，该方法校正效果明显，
对于包含页眉页脚线的扭曲文档图像有较好的校正效果，校正后OCR 识别率大幅度提高。

关键词: 计算机应用, 扭曲文档, 页眉页脚线, 等比距离, 图像校正

Abstract: The recognition rate of OCR (optical character recognition) is low because of the warped
document images. For those warped document images with header and footer lines, a fast method is
proposed to increase the rate of OCR in this paper. Firstly, the location of the header line is detected
and restored in the document image. Then the distance of the line moving upward or downward is
calculated based on geometric algorithm. After that, the image is scanned using the distance as
parameters and the distance that every target pixel needs to remove is calculated. At the same time, all
pixel are removed in order to restructure the image and then a well corrected image is obtained.
Experiments demonstrated that this correcting method was efficient. The OCR rate of warped
document image with header line could be significantly improved.

Key words: computer application, warped document, header and footer line, geometric distance;
image correct