基于局部离群因子和波动阈值的古籍版面图像分析方法

古籍版面图像结构复杂,对其进行有效、准确的分析是实现古籍汉字识别与检索的前提和基础。对古籍汉字版面分析的关键问题展开研究,在对古籍版面特点进行分析与归纳的基础上,提出基于LOF和波动阈值的古籍版面分析方法。首先,采用基于LOF的分类算法对古籍版面图像投影分割后的区域进行分类,确定存在分割问题的候选混合区域；然后,利用波动阈值对候选混合区域中的文字与框线粘连部分进行分割；最后,确定古籍版面中的文字区域并输出。实验结果表明,该算法能够有效地分离古籍文字区域和框线区域,版面分类和分割准确率分别为87.02%和78.69%。

Layout Images Analysis for Ancient Chinese Books Based on LOF and Wave Threshold

It is the premise and basis of recognition and retrieval of ancient Chinese character for realizing automatic analysis of ancient Chinese layout images which is more difficulty than that of the normal printed layout images because of their complex structure. On the basis of the analysis and generalization of layout characteristics of ancient books, the paper puts forward a layout analysis method of ancient Chinese layout images based on LOF and wave threshold. Firstly, LOF-based classification algorithm was used to classify the projected segmentation regions of ancient book layout images, and the candidate mixed regions with segmentation problems were determined. Then, the adhesion parts of text and frame lines in the candidate mixed regions were segmented by using the wave threshold. Finally, the text regions in the ancient book layout were determined and output. The experimental results show that the proposed algorithm can effectively separate the text area and the frame line areas in ancient Chinese layout images, and the layout classification and segmentation accuracy are 87.02% and 78.69% respectively.