基于改进Hough变换文本图像倾斜校正方法_周冠玮.pdfVIP

第 27卷第 7期 计算机应用 Vol. 27 No. 7 2007年 7月 Co puterApplications July 2007 : 1001- 9081( 2007) 07- 1813- 04 Hough 周冠玮,平西建,程 娟 (信息工程大学信息工程学院, 郑州450002) (zgw_text@ yahoo. co . cn) :文本图像在扫描输入时产生的倾斜 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 页面倾斜的快速校正实验结果表明, 该算法实 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 :文本图像; Hough变换;数学形态学; 游程平滑;变分辨率图像 : TP391. 41 : A Skew detection and correction m ethod of docum en t im ages based on im proved H ough tran sform ZHOU Guan-we,i PING X-ijian, CHENG Juan (Co llege of Inf orma tion Eng ineering, Inf ormation Eng ineering University, Zhengzhou H enan 450002, China) A bstract: The skew of docu ent i ages originated fro scanning will bring uch trouble to page seg entation and optical character recognition (OCR). A lthough traditional standardHough transfor is insensitive to noise and discontinuous, it has li itation in application because of huge calculation. A skew detection and correction of docu ent i ages based on i proved Hough transfor was presented. Text-line direction extraction in ult-i resolution i age and reasonable voting threshold selection were introduced to reduce the affect brought by i age skew. Experi ental results show that this ethod is effective and accurate in skew detection and correction of docu ent i ages. K ey w ords: docu ent i age; Hough transfor ; orphology; run-length s oothing; ulti-resolution i age , 0 引言