2.765

2022影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于条件生成对抗网络的书法字笔画分割

张巍 张筱 万永菁

张巍, 张筱, 万永菁. 基于条件生成对抗网络的书法字笔画分割. 自动化学报, 2022, 48(7): 1861−1868 doi: 10.16383/j.aas.c190141
引用本文: 张巍, 张筱, 万永菁. 基于条件生成对抗网络的书法字笔画分割. 自动化学报, 2022, 48(7): 1861−1868 doi: 10.16383/j.aas.c190141
Zhang Wei, Zhang Xiao, Wan Yong-Jing. Stroke segmentation of calligraphy based on conditional generative adversarial network. Acta Automatica Sinica, 2022, 48(7): 1861−1868 doi: 10.16383/j.aas.c190141
Citation: Zhang Wei, Zhang Xiao, Wan Yong-Jing. Stroke segmentation of calligraphy based on conditional generative adversarial network. Acta Automatica Sinica, 2022, 48(7): 1861−1868 doi: 10.16383/j.aas.c190141

基于条件生成对抗网络的书法字笔画分割

doi: 10.16383/j.aas.c190141
基金项目: 国家自然科学基金(61872143)资助
详细信息
    作者简介:

    张巍:华东理工大学信息科学与技术学院硕士研究生. 主要研究方向为数字图像处理. E-mail: johnwayne1995@163.com

    张筱:华东理工大学信息科学与技术学院硕士研究生. 主要研究方向为模式识别. E-mail: zhangxiaoecust17@163.com

    万永菁:华东理工大学信息科学与技术学院教授. 2008年获得华东理工大学检测技术与自动化装置专业博士学位. 主要研究方向为智能信息处理. 本文通信作者. E-mail: wanyongjing@ecust.edu.cn

Stroke Segmentation of Calligraphy Based on Conditional Generative Adversarial Network

Funds: Supported by National Natural Science Foundation of China (61872143)
More Information
    Author Bio:

    ZHANG Wei Master student at the College of Information Sciences and Technology, East China University of Science and Technology. His main research interest is digital image processing

    ZHANG Xiao Master student at the College of Information Sciences and Technology, East China University of Science and Technology. Her main research interest is pattern recognition

    WAN Yong-Jing Professor at the College of Information Sciences and Technology, East China University of Science and Technology. She received her Ph.D. degree in detection technology and automatic equipment from East China University of Science and Technology in 2008. Her main research interest is intelligent information processing. Corresponding author of this paper

  • 摘要: 毛笔书法作为中华传统艺术的精华, 需要在新的时代背景下继续传承和发扬. 书法字是以笔画为基本单元组成的复杂图形, 如果要分析书法结构, 笔画分割是首要的步骤. 传统的笔画分割方法主要利用细化法从汉字骨架上提取特征点, 分析交叉区域的子笔画拓扑结构关系来分割笔画. 本文分析了传统笔画分割基于底层特征拆分笔画的局限性, 利用条件生成对抗网络(Conditional generative adversarial network, CGAN)的对抗学习机制直接分割笔画, 使提取笔画从先细化再分割改进为直接分割. 该方法能有效提取出精确的笔画, 得到的高层语义特征和保留完整信息的单个笔画利于后续对书法轮廓和结构的评价.
  • 图  1  CGAN基本框架

    Fig.  1  Basic framework of CGAN

    图  2  生成器网络结构

    Fig.  2  Network structure of generator

    图  3  判别器网络结构

    Fig.  3  Network structure of discriminator

    图  4  生成器训练过程

    Fig.  4  Generator training process

    图  5  判别器训练过程

    Fig.  5  Discriminator training process

    图  6  测试图像

    Fig.  6  Test image

    图  7  模型训练不同代数的结果

    Fig.  7  Model training results of different epoch

    图  8  损失函数在训练过程中的变化

    Fig.  8  Change of loss function during training

    图  9  5张典型测试图像分割结果

    Fig.  9  Five typical test image segmentation results

    图  10  传统算法骨架法流程

    Fig.  10  Traditional algorithm skeleton method flow

    图  11  本文算法流程

    Fig.  11  The algorithm flow

    图  12  传统算法(上)与本文算法(下)骨架对比

    Fig.  12  Traditional algorithm (top) and the algorithm of this paper (bottom) extract skeleton comparison

    图  13  保留高层语义的两个笔画

    Fig.  13  Two strokes of high-level semantics

    图  14  细化后的两个笔画

    Fig.  14  Two strokes after thining

    表  1  笔画分割的性能

    Table  1  Performance of stroke segmentation

    笔画12345678910111213
    AC0.99960.9976 0.9988 0.9994 0.9996 0.9996 0.9986 0.9991 0.9991 0.9967 0.9992 0.9986 0.9983
    F1 0.9592 0.9435 0.9604 0.9397 0.9710 0.9663 0.95190.93120.9610 0.9583 0.9483 0.9307 0.9572
    下载: 导出CSV
  • [1] 郭晨. 基于图像处理技术的手写体汉字特征分析的研究 [硕士学位论文]. 天津科技大学, 中国, 2010.

    Guo Chen. Research on Character Analysis of Handwritten Chinese Characters Based on Image Processing Technology [Master thesis]. Tianjin University of Science and Technology, China, 2010.
    [2] 李凡. 基于改进K段主曲线算法的图像骨架提取 [硕士学位论文]. 大连海事大学, 中国, 2016.

    Li Fan. Image Skeleton Extraction Based on Improved K-segment Main Curve Algorithm [Master thesis]. Dalian Maritime University, China, 2016.
    [3] 阳平, 娄海涛, 胡正坤. 一种基于骨架的篆字笔划分割方法. 计算机科学, 2013, 40(2):297-300 doi: 10.3969/j.issn.1002-137X.2013.02.066

    Yang Ping, Yan Haitao, Hu Zhengkun. A Skeleton-based Segmentation Method for Scratch Strokes. Computer Science, 2013, 40(2): 297-300(in Chinese) doi: 10.3969/j.issn.1002-137X.2013.02.066
    [4] 苗晋诚. 基于骨架化、骨架划分获取书法汉字结构特征的方法. 昆明理工大学学报:理工版, 2008, 33(3):53-61

    Miao Jincheng. A method for obtaining the structural features of Chinese characters based on skeletonization and skeleton division. Journal of Kunming University of Science and Technology: Science and Engineering Edition, 2008, 33(3): 53-61(in Chinese)
    [5] 章夏芬, 刘佳岩. 用爬虫法提取书法笔画. 计算机辅助设计与图形学学报, 2016, 28(02):301-309. doi: 10.3969/j.issn.1003-9775.2016.02.013

    Zhang Xiafen, Liu Jiayan. Extraction of Calligraphy Strokes by Reptile Method. Journal of Computer-Aided Design and Computer Graphics, 2016, 28(02):301-309.(in Chinese) doi: 10.3969/j.issn.1003-9775.2016.02.013
    [6] 程立, 王江晴, 李波, 田微, 朱宗晓, 魏红昀, 刘赛. 基于轮廓的汉字笔画分离算法. 计算机科学, 2013, 40(07):307-311. doi: 10.3969/j.issn.1002-137X.2013.07.069

    Cheng Li, Wang Jiangqing, Li Bo, Tian Wei, Zhu Zongxiao, Wei Hongwei, Liu Sai. Algorithm for Separation of Chinese Character Strokes Based on Contours. Computer Science, 2013, 40(07): 307-311.(in Chinese) doi: 10.3969/j.issn.1002-137X.2013.07.069
    [7] 曹忠升, 苏哲文, 王元珍, 熊鹏. 基于模糊区域检测的手写汉字笔画提取方法. 中国图象图形学报, 2009, 14(11):2341-2348. doi: 10.11834/jig.20091124

    Cao Zhongsheng, Su Zhewen, Wang Yuanzhen, Xiong Peng. A method for extracting handwritten Chinese characters based on fuzzy region detection. Chinese Journal of Image and Graphics, 2009, 14(11): 2341-2348.(in Chinese) doi: 10.11834/jig.20091124
    [8] 陈睿, 唐雁, 邱玉辉. 基于笔画段分割和组合的汉字笔画提取模型. 计算机科学, 2003(10):74-77. doi: 10.3969/j.issn.1002-137X.2003.10.020

    Chen Rui, Tang Yan, Qiu Yuhui. Extraction model of Chinese strokes based on segmentation and combination of stroke segments. Computer Science, 2003(10):74-77.(in Chinese) doi: 10.3969/j.issn.1002-137X.2003.10.020
    [9] Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets. In: Proceedings of the 27th Annual Conference on Advances in Neural Information Processing Systems (NeurIPS), Montreal, Canada: NIPS, 2014. 2672−2680
    [10] 王坤峰, 苟超, 段艳杰, 林懿伦, 郑心湖, 王飞跃. 生成式对抗网络GAN的研究进展与展望. 自动化学报, 2017, 43(03):321-332.

    Wang Kunfeng, Yan Chao, Duan Yanjie, Lin Yulun, Zheng Xinhu, Wang Feiyue. Research progress and prospects of generatival adversarial network GAN. Acta Automatica Sinica, 2017, 43(03): 321-332.(in Chinese)
    [11] Isola P, Zhu J Y, Zhou T H, Efros A A. Image-to-image translation with conditional adversarial networks. In: Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA: IEEE, 2017. 1125−1134
    [12] Yu J H, Lin Z, Yang J M, Shen X H, Lu Xin, Huang T S. Generative image inpainting with contextual attention. In: Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, USA: IEEE, 2018. 5505−5514
    [13] 张毅锋, 刘袁, 蒋程, 程旭. 用于超分辨率重建的深度网络递进学习方法. 自动化学报, 2020, 46(2): 274−282

    Zhang Yi-Feng, Liu Yuan, Jiang Cheng, Cheng Xu. A deep network progressive learning method for super-resolution reconstruction. Acta Automatica Sinica, 2020, 46(2): 274−282
    [14] Al-Amri S S, Kalyankar N V. Image segmentation by using threshold techniques. arXiv preprint arXiv: 1005.4020, 2010.
    [15] Kang J, Kim S, Oh T J, Chung M J. Moving region segmentation using sparse motion cue from a moving camera. Intelligent Autonomous Systems 12, 2013, 193: 257−264
    [16] Gaur P, Tiwari S. Recognition of 2D barcode images using edge detection and morphological operation. International Journal of Computer Science and Mobile Computing, 2014, 3(4): 1277-1282.
    [17] 刘松涛, 殷福亮. 基于图割的图像分割方法及其新进展. 自动化学报, 2012, 38(06):911-922. doi: 10.3724/SP.J.1004.2012.00911

    Liu Songtao, Yin Fuliang. Image segmentation method based on graph cut and its new progress. Acta Automatica Sinica, 2012, 38(06): 911-922.(in Chinese) doi: 10.3724/SP.J.1004.2012.00911
    [18] Mirza M, Osindero S. Conditional generative adversarial nets. arXiv preprint arXiv: 1411.1784, 2014.
    [19] 蒋芸, 谭宁. 基于条件深度卷积生成对抗网络的视网膜血管分割. 自动化学报, 2021, 47(1): 136−147

    Jiang Yun, Tan Ning. Retinal vascular segmentation based on conditional deep convolution to generatival adversarial network. Acta Automatica Sinica, 2021, 47(1): 136−147
    [20] Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation. In: Proceedings of the 2015 International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015. 234−241
    [21] 杜雪莹. 中国书法 AI 的研究与应用 [硕士学位论文]. 浙江大学, 中国, 2018.

    Du Xue-Ying. Research and Application of Chinese Calligraphy AI [Master thesis]. Zhejiang University, China, 2018.
    [22] Hu M K. Visual pattern recognition by moment invariants. IRE transactions on information theory, 1962, 8(2): 179-187. doi: 10.1109/TIT.1962.1057692
    [23] Zhang Junsong, Yu Jinhui, Mao Guohong, Ye Xiuzi. Denoising of Chinese calligraphy tablet images based on run-length statistics and structure characteristic of character strokes. Journal of Zhejiang University-Science A, 2006, 7(7): 1178-1186. doi: 10.1631/jzus.2006.A1178
    [24] Xu Songhua, Lau F C M, Cheung W K, et al. Automatic generation of artistic Chinese calligraphy. IEEE Intelligent Systems, 2005, 20(3): 32-39. doi: 10.1109/MIS.2005.41
    [25] 张福成. 基于卷积神经网络的书法风格识别的研究 [硕士学位论文]. 西安理工大学, 中国, 2018.

    Zhang Fu-Cheng. Research on Calligraphy Style Recognition Based on Convolutional Neural Network [Master thesis]. Xi'an University of Technology, China, 2018.
  • 加载中
图(14) / 表(1)
计量
  • 文章访问数:  643
  • HTML全文浏览量:  377
  • PDF下载量:  214
  • 被引次数: 0
出版历程
  • 收稿日期:  2019-03-07
  • 录用日期:  2019-06-02
  • 网络出版日期:  2022-06-14
  • 刊出日期:  2022-07-01

目录

    /

    返回文章
    返回