基于改进Hough变换的文本图像倾斜校正方法  

Skew detection and correction method of document images based on improved Hough transform

作　　者：周冠玮[1] 平西建[1] 程娟[1] 

机构地区：[1]信息工程大学信息工程学院,郑州450002

出　　处：《计算机应用》2007年第7期1813-1816,共4页journal of Computer Applications

摘　　要：文本图像在扫描输入时产生的倾斜现象会对后续的页面分割及光学字符识别(OCR)处理产生很大的影响,而传统的标准Hough变换虽然具有对噪声不敏感,不依赖于直线连续性的优点,但由于计算量偏大,速度慢,在实用时有较大的局限性。提出一种基于改进的Hough变换的文本图像倾斜校正方法,通过在变分辨率图像中采用不同的文本方向提取算法,及选择合理投票门限等改进Hough变换的措施,减小了由图像区域及文字笔画粗细所产生的对倾角判定的不利影响,并使用基于偏移值的方法实现页面倾斜的快速校正。实验结果表明,该算法实现了大范围高精度的文本图像倾角的快速检测,具有较强的实用性。The skew of document images originated from scanning will bring much trouble to page segmentation and optical character recognition （OCR）. Although traditional standard Hough transform is insensitive to noise and discontinuous, it has limitation in application because of huge calculation. A skew detection and correction of document images based on improved Hough transform was presented. Text-line direction extraction in multi-resolution image and reasonable voting threshold selection were introduced to reduce the affect brought by image skew. Experimental results show that this method is effective and accurate in skew detection and correction of document images.

关 键 词：文本图像 HOUGH变换 数学形态学 游程平滑 变分辨率图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]