一种基于连通域的蒙古文文档图像版面分析方法  

A Method of Layout Analysis for Mongolian Document Images Based on Connected Components

作　　者：魏宏喜[1] 高光来[1] 

机构地区：[1]内蒙古大学计算机学院,呼和浩特010021

出　　处：《内蒙古大学学报（自然科学版）》2007年第5期586-590,共5页Journal of Inner Mongolia University：Natural Science Edition

基　　金：国家自然科学基金资助项目(批准号69965001)

摘　　要：版面分析是一个将文本页面图像分割成不同区域,并标定区域类型(如文字、图片、表格等)的过程,与字符识别具有同等重要的地位.提出了一种基于连通域的蒙古文版面分析方法,它提取文档图像中所有连通域,根据连通域的大小进行聚类,从而可以得到文字连通域和非文字连通域,达到分割版面的目的.实验证明,该算法能够对蒙古文书籍版面进行准确的分析.Layout analysis is a process that a document image is segmented into different areas and the areas should be classified. It is as important as the character recognition. A new layout analysis method for the Mongolian document images was proposed based on the connected components analysis. All the connected components of a document image are searched by the pixel labeling. Then, they are clustered by their size. Thereby, many connected components of character and non-character can be achieved separately. Experiment shows that the method is suitable for the layout of Mongolian books.

关 键 词：蒙古文文档图像 版面分析 自底向上法 自顶向下法 连通域 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]