摘要
碑刻，泛指刻石文字或图案。碑刻在中国的历史文化传承中占有重要的地位， 是我国古代文化研究的有力的实物实证。由于自然环境和人为损害，碑刻的保护
与研究都面临巨大的挑战。传统的拓印等技术远远无法满足文化保护和传播的要 求，因此需要现代信息技术来更好地记录、处理和展示碑刻。
本文的目的是获取清晰完整的文字轮廓，关于碑文图像的文字提取的几个关键 问题有：图像预处理、笔划提取，轮廓恢复等。首先，本文介绍图像预处理内容， 主要包括：图像去噪，碑文图像单字分割、文字基本骨架提取等。预处理后，碑 文图像中的文字以其骨架的形式表现。
其次，通过分析碑文中汉字结构特点，给出了一种基于汉字横竖撇捺笔划的轮 廓提取技术，然后针对笔划提取过程中遇到的文字轮廓缺失问题，提出了一种简 单的图像融合方法，对文字轮廓缺失部分进行补充融合，得到完整清晰的文字。
接着，在碑文图像的文字展示部分，本文通过 VRML 技术构建一个三维世界 的博物馆场景模型。在这里有两方面的工作，第一展示博物馆的建筑，第二在虚 拟展厅建立石碑模型，将上面工作获取的文字嵌入到石碑模型进行三维展示。
通过试验表明，本文提供的方法可以有效地获取绝大多数碑文图像中的文字信 息。但是在另一方面，由于笔划提取算法较为简单，在某些复杂情况下笔划提取 过程需要人工干预。
关键字：碑刻、文字、提取、轮廓恢复
Abstract
Inscriptions refer stele texts or patterns, which occupy an important position in the heritage of Chinese history and culture. And they are strongly physical evidences of our ancient culture. Due to natural environments and serious human damage, the protection and research of inscriptions suffer enormous challenges. Because traditional rubbing technologies are unable to meet the requirements of the preservation and dissemination of culture, modern information technology is demanded for better recording, processing and displaying inscriptions.
The purpose of this paper is to obtain the outline of the text clearly and completely which includes several key issues: image pre-processing, strokes extraction, contour restoration. Firstly, the content introduces the image pre-processing including: image noise-reduction, word segmentation, basic skeleton extraction. After image pre-processing, the text of the inscription image is expressed by its basic skeleton.
Secondly, through the analysis of the characteristics of Chinese characters on inscriptions, this content puts forward a contour extraction based on four kinds of Chinese strokes which are horizontal stroke, vertical stroke, Write-stroke and flick stroke. Against the problem of missed contours, this paper gives a simple image fusion method which restores the missed contours, so that we could obtain the complete and clear text.
Then, at the
计算机应用技术数字展示 基于数字图像的碑文文字提取及三维展示-计算机应用技术专业论文.docx...
