计算机工程与设计
計算機工程與設計
계산궤공정여설계
COMPUTER ENGINEERING AND DESIGN
2015年
5期
1251-1255
,共5页
中文文本图像%扭曲图像%连通域%文字切分%就近聚合
中文文本圖像%扭麯圖像%連通域%文字切分%就近聚閤
중문문본도상%뉴곡도상%련통역%문자절분%취근취합
Chinese document image%warped image%connected components%character segmentation%nearest aggregation
针对扭曲中文文本图像文字识别率不理想这一问题,提出一种基于连通域的文本图像快速扭曲校正方法。根据汉字结构特征合并连通域,实现切分文字;利用就近聚合文字的方法定位文本行,按行垂直校正每个文字位置,获得被校正的图像。实验结果表明,该方法校正速度快,对严重扭曲的中文文本图像能取得较好的校正效果,校正后图像的OCR识别率明显提高。
針對扭麯中文文本圖像文字識彆率不理想這一問題,提齣一種基于連通域的文本圖像快速扭麯校正方法。根據漢字結構特徵閤併連通域,實現切分文字;利用就近聚閤文字的方法定位文本行,按行垂直校正每箇文字位置,穫得被校正的圖像。實驗結果錶明,該方法校正速度快,對嚴重扭麯的中文文本圖像能取得較好的校正效果,校正後圖像的OCR識彆率明顯提高。
침대뉴곡중문문본도상문자식별솔불이상저일문제,제출일충기우련통역적문본도상쾌속뉴곡교정방법。근거한자결구특정합병련통역,실현절분문자;이용취근취합문자적방법정위문본행,안행수직교정매개문자위치,획득피교정적도상。실험결과표명,해방법교정속도쾌,대엄중뉴곡적중문문본도상능취득교호적교정효과,교정후도상적OCR식별솔명현제고。
Character recognition rate of OCR (optical character recognition)processing is not satisfactory for warped Chinese document image.To resolve this problem,a fast distortion correcting method based on connected components was proposed. First,the connected components were combined together according to the Chinese character structure characteristics.Next,the Chinese characters were segmented one by one according to the combined connected components.After that,the text lines were identified based on the nearest aggregation method.Then,the vertical positions of the segmented characters were corrected ac-cording to every text line.As a result,a well corrected document image was obtained.Experimental results demonstrate that this correcting method is fast and can segment the Chinese character accurately.The OCR rate of the corrected images can be sig-nificantly improved.Even for the obviously distorted Chinese document images,this method can achieve better results.