CASIA OpenIR  > 毕业生  > 博士学位论文
中国水墨作品数字化创作重构研究
唐帆
Subtype博士
Thesis Advisor胡包钢 ; 董未名
2019-05
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword水墨图像 过程重构 知识驱动 绘画分析 动画
Abstract

作为我国最具民族特色的艺术瑰宝与文化符号之一,水墨作品长久以来深受各地人民欢迎。本文研究如何利用数字化技术进行水墨作品分析与创作重构。通过针对水墨作品设计各种描述特征,本文研究了水墨画创作过程动态重构与书法作品重构两类问题。水墨画创作过程重构旨在通过分析静态画作揭示其创作时的绘制过程,并设计渲染方法进行动态绘制。该工作重点关注如何获得与水墨画绘制相关的领域知识,并将这些知识进行抽象化、数字化,用于指导重构工作相关算法设计。书法作品重构旨在利用已有书法作品,使用数据驱动方法挖掘不同形态字体搭配关系,生成新书法作品。针对水墨作品展开的相关研究有助于中国水墨作品的推广与普及。本文通过结合绘画机理研究水墨作品笔画结构分析、笔触语义特征描述方法,并以此为基础展开水墨画创作过程动态重构与书法作品生成。

针对画作内容单一的水墨画作品,提出基于骨架点结构的绘制顺序分析方法。在笔画结构建模基础上,对静态图像进行笔画分割,构建物体拓扑结构的骨架点表达,提出基于结构图遍历的绘制顺序估计方法。针对画作内容复杂的水墨画作品,提出基于多阶段绘制先验的水墨画绘制顺序分析与动态呈现方法。通过多种途径调研分析水墨画绘制先验知识,并根据先验知识类型构建对应的抽象表达,进一步构造优化函数。针对该优化问题,探讨使用自然选择策略进行求解方法。而后,通过对笔触形状分析建模,提出基于椭圆模型的绘制过程动态呈现方法,使水墨画绘制过程按一定节奏、韵律呈现。针对中国传统艺术图像中的书法作品,提出基于自适应字体大小学习策略的书法作品数字化重构方法。本文构建了包含多种书写风格的文字图像数据库,对库中的每个字符进行处理,设计合理的特征对字符图像进行表示,而后通利用学习状态概率转移矩阵挖掘不同文字相互搭配时的大小关系,完成自适应字体大小书法作品数字化重构。

 此外,针对中国水墨艺术图像创作重构问题研究,在研究过程中本文构建的相关应用、收集的素材(水墨画、书法作品)将全面开放用于进一步的研究。

Other Abstract

As one of the most important gems and symbols of Chinese culture, ink-wash artworks have long been popular among the public. This work focuses on two kinds of Chinese Ink-wash artworks reconstruction using digital technology: animated reconstruction of Chinese brush paintings and optimal character composing for Chinese calligraphic artwork. Animated reconstruction of Chinese brush paintings presents a knowledge-driven method for reconstructing the drawing process of Chinese brush paintings. As for optimal character composing for Chinese calligraphic artwork, we develop a data-driven framework to optimally compose multiple characters picked from a database to form a Chinese Calligraphic artwork. Based on the painting mechanism, this work studies the structure of strokes in ink paintings and proposes the analysis of stroke features and then carries out the dynamic reconstruction of ink-wash painting process and the generation of calligraphy.

For ink-wash paintings with single theme, this work proposes a drawing order analysis method based on skeleton point structure. Based on the modeling of stroke structure, we construct the skeletal point expression of the object topology and propose the method of drawing order estimation based on structure graph traversal. For ink-wash paintings with complex content, this work proposes a drawing order analysis method based on a multi-stage painting prior. We map key principles of drawing composition to our computational framework that first organizes the strokes in three different stage.

Then we optimize stroke ordering with natural evolution strategies. For Chinese calligraphy, we propose a data-driven framework to optimally compose multiple characters picked from a database to form a Chinese Calligraphic artwork. Our system can automatically generate a calligraphy in the style of a specific calligrapher. We focus on the adjustment of character sizes in an artwork and formulate the composing of characters as the mining and application of  transition probability matrix. 

Furthermore, for the analysis of computer reconstruction of Chinese ink-wash artworks, a series of applications are built, various kinds of materials are collected. All these application and materials will be released for further research.

Pages110
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/23666
Collection毕业生_博士学位论文
Recommended Citation
GB/T 7714
唐帆. 中国水墨作品数字化创作重构研究[D]. 中国科学院自动化研究所. 中国科学院大学,2019.
Files in This Item:
File Name/Size DocType Version Access License
唐帆_毕业论文.pdf(17508KB)学位论文 暂不开放CC BY-NC-SAApplication Full Text
v6.1.mp4(19058KB)影音 暂不开放ODC BYApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[唐帆]'s Articles
Baidu academic
Similar articles in Baidu academic
[唐帆]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[唐帆]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.