摘要
由于传统书法作品仅包含少量的汉字字体,因此手工绘制一套完整且高质量书法家字体耗时耗力。为了实现书法字体的自动生成,提出一种基于残差单元的生成对抗网络书法字体生成方法,该方法使用基于残差单元的生成器学习书法字体的特征,构建三要素损失函数进行网络训练,通过生成器和鉴别器的对抗训练,对生成的字体进行真假判断,实现汉字书法字体自动生成。实验构建了颜真卿颜体、欧阳询楷体、启功简体和文征明小楷4种书法小样本字体库数据集,实验表明,本文方法不仅能够生成逼真的书法字体,还具有生成完整字体库的能力。
It takes a lot of time and effort to draw a complete and high-quality calligrapher′s font by manual due to traditional calligraphy works only has a small number of Chinese characters.In order to realize the automatic generation of calligraphy fonts,a calligraphy fonts generation method based on residual generative adversarial network is proposed in this paper.Firstly,the characteristics of calligraphy fonts are learnt by the generator based on residual units.Then,the loss function based on three elements is constructed to train the font generation network.Finally,the adversarial training between the generator and the discriminator is used to judge whether the generated font is true or false,and realize the automatic generation of Chinese calligraphy font.Experiments have performed on four kinds of calligraphy datasets:Yan Zhenqing′s Yan Ti,Ouyang Xun′s Kai Ti,Qi Gong′s Jian Ti and Wen Zhengming′s Xiao Kai.The experimental results on four datasets indicate that the proposed method achieves good font generation result,which can generate realistic calligraphy fonts and has the ability to generate a complete font library.
作者
赵静
吴晓军
杨红红
苏玉萍
ZHAO Jing;WU Xiaojun;YANG Honghong;SU Yuping(School of Computer Science,Shaanxi Normal University,Xi'an 710119,Shaanxi,China)
出处
《陕西师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2021年第5期85-93,共9页
Journal of Shaanxi Normal University：Natural Science Edition
基金
国家重点研发计划(2017YFB1402102)。
关键词
书法艺术
卷积神经网络
生成对抗网络
字体生成
calligraphy art
convolutional neural network
generative adversarial network
font generation
