公开(公告)号：JP2005309608A

公开(公告)日：2005-11-04

申请号：JP2004123277

申请日：2004-04-19

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： KATO MAKOTO ,  ARIMA TOSHIMICHI ,  SUMIKAWA MICHITOSHI

IPC: G06K9/03 ,  G06K9/46 ,  G06K9/50 ,  G06K9/62

CPC classification number: G06K9/50 ,  G06K2209/01

Abstract: PROBLEM TO BE SOLVED: To reduce the burden of an operator by displaying character images similar in letter shape whose typefaces are similar collectively in a confirmation screen where the character images in the same category are arranged, thereby improving the operation efficiency of confirming/correcting recognition results by the operator. SOLUTION: This output mechanism of a character recognition device is provided with: a category classifying part 20 for classifying the image data of characters being the target of the character recognition processing for every character(category) recognized by the character recognition processing; a clustering processing part 30 for calculating featured values related with the shapes of the characters included in the image data in each category classified by the category classifying part 20, and for classifying the image data into one or more clusters based on the featured values; and a picture generating part 50 for generating and displaying the confirmation picture on which the image data are displayed for every cluster classified by the clustering processing part 30. COPYRIGHT: (C)2006,JPO&NCIPI

Abstract translation: 要解决的问题：通过在同一类别的字符图像的确认画面中显示与字体形状相似的字符形状的字符图像来减轻操作者的负担，从而提高确认的操作效率 /操作员校正识别结果。 解决方案：字符识别装置的输出机构具有：类别分类部分20，用于对通过字符识别处理识别的每个字符（类别）对作为字符识别处理的目标的字符的图像数据进行分类; 聚类处理部分30，用于计算与由类别分类部分20分类的每个类别中包括在图像数据中的字符的形状相关的特征值，并且基于特征值将图像数据分类成一个或多个聚类; 以及图像生成部50，用于生成并显示对于通过聚类处理部30分类的每个聚类显示图像数据的确认图像。（C）2006，JPO＆NCIPI

公开(公告)号：JP2002366895A

公开(公告)日：2002-12-20

申请号：JP2001163376

申请日：2001-05-30

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーション

Inventor： SUMIKAWA MICHITOSHI ,  TAKAHASHI HIROYASU

IPC: G06K9/20 ,  G06K9/32 ,  G06Q10/10 ,  G06T7/00 ,  G06T7/60 ,  G06F19/00

CPC classification number: G06K9/2054 ,  G06K2209/01

Abstract: PROBLEM TO BE SOLVED: To provide technology which can stably discriminate a document by generating a stable virtual page mark even for a non-OCR document.
SOLUTION: To detect the virtual page mark, a key segment which can stably be detected in the document is previously defined and on the basis of the key segment, the virtual page mark (circumscribed rectangle) is generated. Redundancy is given to the detection of the key segment and even if the key segment can not be detected because of a stain, absence, etc., an alternative segment is defined to generate the virtual page mark on the basis of the alternative segment. The selected key segment meets conditions; (1) a segment which is thick enough to be tolerant of faintness of the document, (2) a segment which is at a sufficient distance from the circumference of the document and stable against skew and the absence of a document end, and (3) a segment which does not overlap with a fold of the original paper of the document.
COPYRIGHT: (C)2003,JPO

Abstract translation: 要解决的问题：提供即使对于非OCR文档也可以通过生成稳定的虚拟页标记来稳定地区分文档的技术。 解决方案：为了检测虚拟页面标记，先前定义了可以稳定地检测到的文档中的关键段，并且基于关键段，生成虚拟页面标记（外接矩形）。 给予关键段的检测冗余，即使由于污点，缺失等而无法检测到关键段，也可以定义替代段，以便在替代段的基础上生成虚拟页标记。 所选关键段满足条件; （1）足够厚以容忍文件微弱的部分，（2）与文件圆周距离足够远的部分，并且稳定地抵抗歪斜和文档端部的缺失，（3） ）不与文件的原始纸张的折叠重叠的片段。

公开(公告)号：JPH11143986A

公开(公告)日：1999-05-28

申请号：JP28570997

申请日：1997-10-17

Applicant: IBM

Inventor： SUMIKAWA MICHITOSHI ,  TAKAHASHI HIROYASU

IPC: G06K9/20 ,  G06K9/32

Abstract: PROBLEM TO BE SOLVED: To enable the specification of a character frame and character recognition of a business form without a page mark and a reference mark even with a scanner incapable of detecting an edge of the business form and to simultaneously increase the speed of a discrimination processing of a bit map image by comparing images based on a externally contacted rectangle, etc., formed from horizontal segments capable of being detected at high speed. SOLUTION: When the bit map image including the horizontal segments like character frames 323, 327 and ruled lines 301, 303 on the business form is discriminated as in the case that the business form 300 with the black character frame and without the page mark is discriminated and the character is recognized by an OCR, the horizontal segments are extracted as characteristics of the business form, the circumscribed rectangle 350 is formed in an area to be generated from the horizontal segments and the circumscribed rectangle 350 is defined as information to discriminate an estimation standard of a position of the character frame and a kind of the business form. Even the business form without the page mark and the reference mark is recognized by adapting the information to the OCR. In addition, the business form is more exactly discriminated by comparing the extracted horizontal segments themselves with the horizontal segment of a preliminarily registered business form definition body and comparing similarity.