CASIA OpenIR  > 毕业生  > 硕士学位论文
Thesis Advisor刘成林
Degree Grantor中国科学院研究生院
Place of Conferral北京
Keyword笔迹合成 风格变换 骨架提取 点集配准
Abstract 笔迹合成是在获取部分笔迹样本的情况下,对样本进行形状分析,模拟生成类似书写风格字迹的技术,常应用于样本生成、辅助字体设计、个人笔迹模仿等。常见的笔迹合成方法有基于扰动、部件重组、统计模型等方法。
l  实现基于薄板样条插值函数的点集匹配方法,并介绍了点集配准在字符模板匹配、字符变形、笔划拆分等方面的应用;
l  实现了一种骨架化算法,用来从脱机字符图像提取笔划轨迹;
l  提出了一种基于神经网络的风格变换的手写汉字笔迹合成方法,包括笔划点上下文特征信息的提取、变形网络的结构等。为提高笔迹合成的质量,提出了在神经网络训练过程中加入笔划邻近点相对位置信息来约束笔划轨迹点变形的方案。
Other AbstractThe purpose of handwriting synthesis is to generate new samples by simulating the writing style of a small number of written samples. Handwriting synthesis is often used in  samples generation, auxiliary font design, personal handwriting imitation, and so on. The methods of handwriting synthesis can be grouped into three categories: perturbation-based generation, fusion-based generation, statistical model-based generation.
   Based on reviewing the traditional character synthesis methods, we propose a novel learning-based approach to generate personal style handwritten characters by estimating the shape transformation from standard templates to personal style characters. The main works are as follows:
l  We implement the point set matching method based on the Thin Plate Spline for aligning the stroke points between character templates and written characters, to facilitate the estimation of shape transformation;
l  For extracting the stroke trajectory of offline character images, we implement a skeletonization algorithm;
l  We propose a Chinese handwritten  character generation method by style transformation using a neural network, which models the deformation between the stroke points of templates and written characters. To improve the smoothness of generated character strokes we regularize the transformation using the distortion of adjacent points.
   The experimental results show that the proposed algorithm can synthesize high-quality and specific style handwriting when using a small amount of written samples for training.
Subject Area模式识别
Document Type学位论文
Recommended Citation
GB/T 7714
谭必仁. 基于神经网络的中文笔迹合成方法研究[D]. 北京. 中国科学院研究生院,2017.
Files in This Item:
File Name/Size DocType Version Access License
基于神经网络的中文笔迹合成方法研究.pd(1833KB)学位论文 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[谭必仁]'s Articles
Baidu academic
Similar articles in Baidu academic
[谭必仁]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[谭必仁]'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.