基于深度学习的木刻版画图像生成方法
Method for generating woodcut images Based on deep learning
投稿时间:2022-06-10  修订日期:2022-06-10
DOI:
中文关键词: 木刻版画  图像生成  深度学习  神经网络  风格转换
英文关键词: woodcut print  image generation  deep learning  neural network  style transfer
基金项目:
作者单位
郭丽娟 延安大学鲁迅艺术学院 
摘要点击次数: 165
全文下载次数: 0
中文摘要:
      木刻版画有着悠久的历史与极高的艺术价值,它是一项对审美与雕刻技法有极高要求的艺术创作过程,需要专业人士花费大量的时间与精力去完成。如何让机器学习艺术家的木刻版画创作过程与创作风格是一项具有挑战性的研究。目前基于计算机与人工智能的木刻版画图像生成方法多数仅是通过构造硬编码方法去模仿木版画的纹理、色调等模式,缺乏像艺术家一样对创作过程的感知与理解。为了传承木刻版画艺术,利用计算机辅助完成木刻版画创作本文提出一种基于深度神经网络的木版画图像生成方法,可将一副摄影图片转化为给定的某种木版画风格的图像,并创建了一个木版画的图像数据集;借鉴深度学习的风格迁移技术设计出了具有文本提示功能的交互式木版画风格图像的生成算法。实验结果表明本文算法能很好模拟艺术家的创作风格,可生成各种不同自然场景和人物的木刻版画图像,算法运行高效稳定。
英文摘要:
      Woodcut engraving has a long history and high artistic value. It is an artistic creation process with high requirements for aesthetics and carving techniques, which requires professionals to spend a lot of time and energy to complete. It is a challenging research on how to let the machine learn the creative process and style of woodcut prints of artists. At present, most of the image generation methods of woodcut based on computer and artificial intelligence only imitate the texture, tone and other modes of woodcut by constructing hard coding methods, lacking the perception and understanding of the creative process like artists. In order to inherit the woodcut art and complete the woodcut creation with computer assistance, this paper proposes a method of woodcut image generation based on deep neural network, which can transform a photographic picture into an image of a given woodcut style, and create a woodcut image data set; Using the style transfer technology of deep learning for reference, an interactive woodcut style image generation algorithm with text prompt function is designed. The experimental results show that the algorithm can well simulate the artist''s creative style, and can generate woodcut prints of various natural scenes and figures. The algorithm runs efficiently and stably.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭