CASIA OpenIR  > 毕业生  > 硕士学位论文
版面理解及其算法研究
王海琴
学位类型工学硕士
导师戴汝为
1996-06-01
学位授予单位中国科学院自动化研究所
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
摘要本文首先分析了当前国内外在版面分析和版面理解方面的研究现状,综合 评估了现有的主要版面理解算法的优缺点和可行性。在此基础上,系统总结了 作者在硕士研究生学习阶段的研究工作,初步提出了一个新的图象特征函 数一一穿线,并将此函数与基于知识的算法思想结合起来,应用于版面的分析 和理解算法中,独立完成了一个实用的版面分析系统,该系统在实际应用中取 得了良好的效果。 本文的主要工作是: 1. 作为讨论和研究版面理解算法的基础,对文件版面的类型作了一个系统 的划分,从而为评价版面分析和理解算法提供了一个有效的依据和标准。 2. 系统分析了目前普遍使用的两种图象文件特征函数一一投影和游长,提 出了一种兼有两者优点的新的持征函数一一穿线,对穿线的性质和表达能力进 行了全面的分析,并把它应用于版面理解和分析算法中。 3.进一步地将穿线应用于图象预处理方面,提出了基于穿线的倾斜检测和 校正的方法,以及利用穿线进行点噪声识别的方法,从而提高了版面分析算法 的抗倾斜、抗噪声能力。 4 把基于知识的算法思想应用在版面理解和分析系统中,根据文件排版的 特点,特别是中文杂志、报纸等版面的共同特点,总结了一些对版面分析具有 普遍指导意义的规则知识,在版面分析算法中将这些规则作为启发式引导,实 现了基于知识的分析理解算法。 5 将穿线分析的方法和基于知识的思想综合起来,开发了一个具有一定的 理论意义和实用价值的版面理解系统,并取得了较好的分析理解效果。 6. 根据综合集成的思想,对版面分析理解的发展前景作了一些探讨,在基 于知识的版面分析和理解算法的基础上,初步分析了人机一体化系统对未来版 面分析理解算法研究的意义。
其他摘要The thesis first analyzes the national and international development situation of document analysis and understanding, then discusses the strong and weak points of the main algorithm for document analysis. Besides, the feasibility of each algorithm is discussed. On the basis of the discussion, the research program during author's graduate student period is introduced in a systematic way. A new feature of document image named Crosscount is provided. Combining Crosscount with the knowledge-based algorithm, a feasible system for document analysis is designed. It is applied to practical use and the result is satisfied. The main work of the thesis is the following: 1. As the base of discussing and studying the algorithm for document analysis and understanding, the document images is analyzed and classified by their layout structures. The classification provides a foundation for judging and evaluating the document analysis and understanding algorithm. 2. Two frequently-used features of document images named Projection and Run Length is analyzed in chapter two. A new feature named Crosscount combined the two's advantages is provided. The characteristics and description abilities of Crosscount is discussed in details. Its application in document analysis and understanding is also presented in the thesis. 3. Further more, Crosscount is applied to images preprocessing. Two methods for skew detection and correction, and one for dotted noises detection are introduced. Using these methods, the algorithm can improve its skew-resistance and noise-resistance. 4. The idea of the knowledge-based algorithm is applied to document analysis and understanding. According to the characteristics of composition, especially the distinguishing features of Chinese news-paper and journals, some general-objected rules of composition is summarized. Applying these heuristic rules to the document analysis、 a knowledge-based algorithm for document analysis and understanding has been designed and developed. 5. By combining the Crosscount feature and the knowledge-based algorithm, a document understanding system with both theoretic and practical values has been finished. It has got satisfied results. 6. According to the idea of metasynthesis, the development prospect of document analysis and understanding is discussed. On the basis of knowledge-based algorithm for document analysis and understanding, the significance of the man-computer metasynthetic system to the future of document analysis and understanding is presented in the last part of the thesis.
馆藏号XWLW395
其他标识符395
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/7166
专题毕业生_硕士学位论文
推荐引用方式
GB/T 7714
王海琴. 版面理解及其算法研究[D]. 中国科学院自动化研究所. 中国科学院自动化研究所,1996.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[王海琴]的文章
百度学术
百度学术中相似的文章
[王海琴]的文章
必应学术
必应学术中相似的文章
[王海琴]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。