广东工业大学学报 ›› 2019, Vol. 36 ›› Issue (03): 47-55.doi: 10.12052/gdutxb.180162

• 综合研究 • 上一篇    下一篇

基于CGAN网络的二阶段式艺术字体渲染方法

叶武剑1, 高海健1, 翁韶伟1, 高智1, 王善进2, 张春玉3, 刘怡俊1   

  1. 1. 广东工业大学 信息工程学院, 广东 广州 510006;
    2. 东莞理工学院 电子工程与智能化学院, 广东 东莞 523808;
    3. 西藏民族大学 信息工程学院, 陕西 咸阳 712082
  • 收稿日期:2018-11-28 出版日期:2019-05-09 发布日期:2019-04-04
  • 通信作者: 翁韶伟(1980-),女,副教授,博士,主要研究方向为图像可逆水印、信息隐藏、视频篡改等.E-mail:wswweiwei@126.com E-mail:wswweiwei@126.com
  • 作者简介:叶武剑(1987-),男,讲师,博士,主要研究方向为深度学习、计算机视觉、计算机网络等.
  • 基金资助:
    国家自然科学基金资助项目（61872095，61872128，61571139，61201393）；广东省信息安全技术重点实验室开放课题基金资助项目（2017B030314131）；广东省智能信息处理重点实验室、深圳市媒体信息内容安全重点实验室2018年开放基金课题资助项目（ML-2018-03）；西藏自治区自然科学基金资助项目（2016ZR-MZ-01）；广州市珠江科技新星专题项目（2014J2200085）；广东工业大学青年百人项目资助项目（220413548）

A Two-stage Effect Rendering Method for Art Font Based on CGAN Network

Ye Wu-jian1, Gao Hai-jian1, Weng Shao-wei1, Gao Zhi1, Wang Shan-jin2, Zhang Chun-yu3, Liu Yi-jun1   

  1. 1. School of Information Engineering, Guangdong University of Technology, Guangzhou, 510006, China;
    2. School of Electrical Engineering & Intelligentization, Dongguan University of Technology, Dongguan, 523808, China;
    3. School of Information Engineering, Xizang Minzu University, Xianyang, 712082, China
  • Received:2018-11-28 Online:2019-05-09 Published:2019-04-04

HTML

PDF (PC)

2808

摘要： 艺术字体渲染是媒体排版的重要技术之一.如何提供一种高效的艺术字体渲染方法来实现生成艺术字体的特效多样化与清晰化是亟待解决的问题.本文借助条件生成式对抗网络（Conditional Generative Adversarial Networks，CGAN），提出一个包括风格化处理和清晰化处理的二阶段式艺术字体渲染方法，对字体实现高效的特定效果渲染.首先，风格化处理是通过构建风格化网络模型对多样化的字体进行各种不同的2D或3D特效渲染；然后，构建清晰化网络模型对生成的艺术字体图进行清晰化处理，这克服了单一GAN网络生成图模糊的缺陷.实验结果表明，二阶段式艺术字体渲染方法所生成的特效字体的纹理细节较为丰富，不受限于文字骨架，而且字体清晰度也得到较大提升；同时，该方法对字体的特效渲染批量化处理效率也明显提高，具有较强实用价值.

关键词: 艺术字体, 特效渲染, 条件生成式对抗网络, 风格化, 清晰化

Abstract: Art font rendering is one of the important techniques of media typesetting. How to provide an efficient art font rendering method to realize the diversity and clearness of the special effects of generative art fonts is an urgent problem for technicians in this field. With the help of Conditional Generative Adversarial Networks (CGAN), a two-stage art font rendering method that includes style transfer processing and enhancement processing is proposed to achieve high quality rendering of art fonts. First, by style transfer processing, a stylized network model is built to render various 2D or 3D special effects on a variety of input fonts. Then, by enhancement processing, a sharpening network model is built to sharpen the generated art font images, which overcomes the defect of fuzzy image generation in a single GAN network. The experimental results show that, comparing with existing methods, this scheme has a better performance for generating more diverse and more clear special art fonts, whose texture details are relatively rich, and are not limited to the text skeleton. At the same time, this scheme can also improve the efficiency of art font rendering in the batch processing and has a higher practical value.

Key words: art font, effect rendering, conditional generative adversarial networks, style transfer, sharpness

中图分类号: 

  • TP391
[1] YANG S, LIU J, LIAN Z, et al. Awesome typography:statistics-based text effects transfer[J]. IEEE Computer Vision and Pattern Recognition, 2017:7464-7473
[2] BARNES C, SHECHTMAN E, GOLDMAN D B, et al. The generalized patchmatch correspondence algorithm[C]//European Conference on Computer Vision. Berlin, Heidelberg:Springer 2010:29-43.
[3] BARNES C, SHECHTMAN E, FINKELSTEIN A, et al. PatchMatch:a randomized correspondence algorithm for structural image editing[J]. ACM Transactions on Graphics (ToG), 2009, 28(3):24.
[4] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]//International Conference on Neural Information Processing Systems. Canada, Montreal:MIT Press, 2014:2672-2680.
[5] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein generative adversarial networks[C]//International Conference on Machine Learning. Sydney,Australia:ICML,2017:214-223
[6] CHANG J, SCHERER S. Learning representations of emotional speech with deep convolutional generative adversarial networks[C]//2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). New Orleans:IEEE, 2017, 2746-2750.
[7] ISOLA P, ZHU J Y, ZHOU T, et al. Image-to-image translation with conditional adversarial networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Honolulu, Hawaii:IEEE, 2017, 1125-1134.
[8] ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE International Conference on Computer Vision. Italy, Venice:IEEE, 2017:2223-2232.
[9] ZHU J Y, KRÄHENBÜHL, PHILIPP, SHECHTMAN E, et al. Generative visual manipulation on the natural image manifold[J]. 浙江大学学报(英文版)(c辑:计算机与电子), 2016, 15(4):312-320
[10] 曾碧, 任万灵, 陈云华. 基于CycleGAN的非配对人脸图片光照归一化方法[J]. 广东工业大学学报, 2018, 35(5):11-19 ZENG B, REN W L, CHEN Y H. An unpaired face illumination normalization method based on cycleGAN[J]. Journal of Guangdong University of Technology, 2018, 35(5):11-19
[11] LI Y, FANG C, YANG J, et al. Universal style transfer via feature transforms[C]//Conference and Workshop on Neural Information Processing Systems. Long Beach:NIPS, 2017:386-396.
[12] JOHNSON J, ALAHI A, FEI-FEI L. Perceptual losses for real-time style transfer and super-resolution[C]//European conference on computer vision. Amsterdam:Springer, 2016:694-711.
[13] GATYS L A, ECKER A S, BETHGE M. Image style transfer using convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas:IEEE, 2016:2414-2423.
[14] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Boston, Massachusetts:IEEE, 2016:770-778.
[15] SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas:IEEE, 2016:2818-2826.
[16] SZEGEDY C, IOFFE S, VANHOUCKE V, et al. Inception-v4, inception-resnet and the impact of residual connections on learning[C]//Thirty-First AAAI Conference on Artificial Intelligence. San Francisco, California. 2017:AAAI, 4278-4284.
[1] 谢国波, 林立, 林志毅, 贺笛轩, 文刚. 基于YOLOv4-MP的绝缘子爆裂缺陷检测方法[J]. 广东工业大学学报, 2023, 40(02): 15-21.
[2] 陈靖宇, 吕毅. 基于脉冲神经网络的冷链制冷机结霜检测方法[J]. 广东工业大学学报, 2023, 40(01): 29-38.
[3] 叶文权, 李斯, 凌捷. 基于多级残差U-Net的稀疏SPECT图像重建[J]. 广东工业大学学报, 2023, 40(01): 61-67.
[4] 邹恒, 高军礼, 张树文, 宋海涛. 围棋机器人落子指引装置的设计与实现[J]. 广东工业大学学报, 2023, 40(01): 77-82,91.
[5] 谢光强, 许浩然, 李杨, 陈广福. 基于多智能体强化学习的社交网络舆情增强一致性方法[J]. 广东工业大学学报, 2022, 39(06): 36-43.
[6] 刘信宏, 苏成悦, 陈静, 徐胜, 罗文骏, 李艺洪, 刘拔. 高分辨率桥梁裂缝图像实时检测[J]. 广东工业大学学报, 2022, 39(06): 73-79.
[7] 熊武, 刘义. 粒子滤波算法在BDS高铁铁轨静态形变监测中的应用研究[J]. 广东工业大学学报, 2022, 39(04): 66-72.
[8] 易闽琦, 刘洪伟, 高鸿铭. 电商平台产品共同购买网络的影响因素研究[J]. 广东工业大学学报, 2022, 39(03): 16-24.
[9] 丘展春, 费伦科, 滕少华, 张巍. 余弦相似度保持的掌纹识别算法[J]. 广东工业大学学报, 2022, 39(03): 55-62.
[10] 郑佳碧, 杨振国, 刘文印. 基于细粒度混杂平衡的营销效果评估方法[J]. 广东工业大学学报, 2022, 39(02): 55-61.
[11] Gary Yen, 栗波, 谢胜利. 地球流体动力学模型恢复的长短期记忆网络渐进优化方法[J]. 广东工业大学学报, 2021, 38(06): 1-8.
[12] 李光程, 赵庆林, 谢侃. 去中心化的数据处理方案设计[J]. 广东工业大学学报, 2021, 38(06): 77-83.
[13] 谢光强, 赵俊伟, 李杨, 许浩然. 基于多集群系统的车辆协同换道控制[J]. 广东工业大学学报, 2021, 38(05): 1-9.
[14] 张巍, 张圳彬. 联合图嵌入与特征加权的无监督特征选择[J]. 广东工业大学学报, 2021, 38(05): 16-23.
[15] 邓杰航, 袁仲鸣, 林好润, 顾国生. 协同超像素和视觉显著性的图像质量评价[J]. 广东工业大学学报, 2021, 38(05): 33-39.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发