公开(公告)号：KR1020040044656A

公开(公告)日：2004-05-31

申请号：KR1020020072754

申请日：2002-11-21

Applicant: 한국전자통신연구원

Inventor： 김계경 ,  정연구 ,  지수영 ,  유원필 ,  이형구 ,  조수현

IPC: G06K9/50

CPC classification number: G06K9/342 ,  G06K2209/01

Abstract: PURPOSE: A handwritten number segmentation method is provided to find non-vertical segmentation lines of handwritten numbers with an arbitrary incline, to definitely segment handwritten numbers and to recognize each number. CONSTITUTION: The method comprises several steps. An image of handwritten numbers with non-vertical segmentation lines is inputted(S1). The image, if it is bent, is smoothed(S2). A profile is extracted from the smoothed image(S3). Candidate segmentation areas, where there can exist numbers, are searched in the profile(S4). Candidate segmentation points are searched in the candidate segmentation areas(S5). Vertical segmentation lines are searched in the candidate segmentation points(S6). Non-vertical segmentation lines with arbitrary inclines are searched in the candidate segmentation points(S7). Attached numbers are segmented by using the vertical and non-vertical segmentation lines(S8).

Abstract translation: 目的：提供手写数字分割方法，用于查找具有任意倾斜度的手写数字的非垂直分割线，以确定分割手写数字并识别每个数字。 构成：该方法包括几个步骤。 输入具有非垂直分割线的手写数字的图像（S1）。 如果图像弯曲，则平滑（S2）。 从平滑图像中提取轮廓（S3）。 在轮廓中搜索候选分割区域（可存在数字）（S4）。 在候选分割区域中搜索候选分割点（S5）。 在候选分割点中搜索垂直分割线（S6）。 在候选分割点中搜索具有任意倾斜的非垂直分割线（S7）。 通过使用垂直和非垂直分割线来分割附加的数字（S8）。

公开(公告)号：KR1020020055454A

公开(公告)日：2002-07-09

申请号：KR1020000083420

申请日：2000-12-28

Applicant: 한국전자통신연구원

Inventor： 지수영 ,  장대근 ,  황영섭 ,  문경애 ,  조수현 ,  정연구

IPC: G06T7/40

CPC classification number: G06K9/00456

Abstract: PURPOSE: A method of interpreting a document image area is provided to extract connected components to group the connected components as tree structures according to spacial relations, and to readjust the components in a text area via separating/combining procedures, thereby efficiently interpreting a document structure. CONSTITUTION: Connected components are analyzed through a reduced document image(61, 62). A tree is generated by an analyzed result of the connected components, to classify the connected components(63, 64). Text factors are grouped according to spacial relations from the classified connected components. A text block is readjusted through separation/combination procedures of the connected components. The step of generating the tree and classifying the connected components comprises the steps as follows. The tree is constructed from types of the connected components. Connected components including tables, frames, and pictures are grouped as independent nodes with text. Connected components within a text block surrounded by margins are grouped. Nodes which are not grouped are classified by areas of the connected components.

Abstract translation: 目的：提供一种解释文档图像区域的方法，以提取连接的组件，以根据空间关系将连接的组件分组为树结构，并通过分离/组合过程重新调整文本区域中的组件，从而有效地解释文档结构 。 构成：通过缩小的文档图像分析连接的组件（61,62）。 通过连接的组件的分析结果生成树，以对连接的组件进行分类（63,64）。 文本因素根据与分类的连通组件的空间关系进行分组。 通过连接组件的分离/组合程序重新调整文本块。 生成树并对连接的组件进行分类的步骤包括以下步骤。 树是由连接的组件的类型构成的。 连接的组件，包括表，框架和图片被分组为具有文本的独立节点。 通过边距包围的文本块中的已连接组件进行分组。 未分组的节点按照连接组件的区域进行分类。