摘要
为解决扭曲文本图片光学字符识别率低的问题,提出基于逆向工程的文档图像扭曲恢复算法。用三维扫描仪采集书本三维模型,运用OpenGL库打开obj文件,依次进行基于书籍平面的正向投影和基于离散点坐标值的拉伸校正,对由厚度造成的书籍文字的扭曲现象进行校正,产生展平书页的逆向效果。实验结果表明,该算法可有效提高OCR识别率。
To tackle the problem of low OCR（optical character recognition）rate caused by distorted document image,a distorted document image restoration algorithm based on reverse engineering was introduced.Three-dimension model of the book was collected,and obj file was opened with the help of OpenGL.After the process of projection and stretch,the distortion caused by the thickness of the book was restored,thus making the book flat in 3Darchitecture.Results of experiment show that the algorithm can flatten the pages in the book effectively and the accuracy rate of OCR is improved.
出处
《计算机工程与设计》
北大核心
2016年第4期964-968,共5页
Computer Engineering and Design
基金
国家自然科学基金重点项目(61371142)
关键词
光学字符识别
逆向工程
三维模型
扭曲校正
投影
optical character recognition
reverse engineering
3D model
distorted image restoration
projection
