多层次可信度指导下的自底向上的版面分析算法  

A BOTTOM-UP LAYOUT ANALYSIS ALGORITHM BASED ON MULTI-LEVEL CONFIDENCE

作　　者：陈明[1] 丁晓青[1] 吴佑寿[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《模式识别与人工智能》2003年第2期198-203,共6页Pattern Recognition and Artificial Intelligence

摘　　要：版面分析是文档数字化中的关键问题。本文首先将版面分析算法分为两类:基于形状的和基于纹理的,然后提出了一个基于多层次基元的版面分析模型,根据此模型,版面分析问题被简化为计算每个层次上的最佳基元划分。我们提出了连接强度和多层次可信度的概念并描述了一个基于多层次可信度的自底向上的算法。此算法具有很强的适应性,可以处理各种类型的样张。实验表明了算法的有效性。Layout analysis is a key problem in the document digitalization. In this paper, layout analysis algorithms are classified to two categories: shape-based method and texture-based method. Then a layout analysis model based on multi-level primitives is proposed and the problem of layout analysis is simplified to compute the best partition on each level. Based on this model, the concept of multi-level confidence is introduced and bottom-up algorithm based on multi-level confidence is described. Analysis under the Multi-level confidence is very flexible for all kinds of documents. Experimental results have proved that the algorithm proposed by this paper is very effective.

关 键 词：文档图像分析 版面分析算法 图像分割 多层次可信度 纹理分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]