云南苗族刺绣数字合成及其应用研究

Automatic generation of artistic illustrations from images is one of the most fundamental and actively studied problems to Non-Photorealistic Rendering (NPR). Nowadays, the application of the NPR research results is related to the network, digital entertainment, advertisement etc. With the development of NPR, computer can transform images into specific artistic styles such as ink painting, Chinese Calligraphy etc., or, more generally in a style other than realism. NPR effects are more engaged than the traditional, photorealism digital images rendering effects. Though there are many art styles have been simulated through NPR methods, there is few techniques to research Miao nation embroidery of Yunnan Provice. So, based on some NPR techniques, the project will simulate the Miao nationality embroidery style art. . Miao embroidery is an art of decorating textiles with needle and thread, and it is a very famous national artistic work in Yunnan and Guizhou Plateau. During the research of our proposal, we will investigate and analyse the real creative process of embroidery, and expose some characters of this art, like embroidery color, embroidery thread, stitch etc. Firstly, the database of Miao embroidery’s texture and patterns will be built, and we also improve texture synthesis method to rich the patterns. Secondly, we simulate the colors of Miao embroidery based on color transfer and illumination model. Further, the most important of Miao embroidery’s character is embroidery process, acupuncture method and multidimensional visual surface, we will build corresponding acupuncture database and propose the digital synthesis algorithm to simulate these creation skills. The proposal will adopt some acceleration and optimization methods to generate final Miao embroidery artistic effects. In addition, the 3D garment model of Miao embroidery and human body model will be built, then the project will extend simulating results to 3D virtual fitting system. . The findings of this project will provide a new NPR artistic field of Miao embroidery and enlarge application fields of NPR. The proposed methods provide a positive role on promotion and protection the cultural heritage of Chinese national minority. The project is expected to publish 10-15 papers in national and international journals and conference, and exploit realtime rendering system about Miao embroidery artistic effects.

在非真实感绘制技术的发展过程中，对不同艺术风格的数字化仿真不断吸引着众多研究者，研究对象和应用领域不断扩展。国内外研究人员对中国工笔画、书法等艺术作品进行了数字化模拟，然而针对云南苗族刺绣艺术作品的数字化模拟并不多见。本项目立足于云南的苗族特色文化，选取国内外享有盛誉的苗绣作品作为研究对象，考察和分析苗绣作品的实际创作过程，揭示苗绣构图、色彩、针法、画面等方面的规律和特点，对其数字化仿真开展研究，重点针对苗绣作品中丰富的纹样和图案、色彩、刺绣技法、立体视觉效果等方面提出相应的数字化模拟算法，同时研究苗绣数字化模拟过程的加速和优化方法。此外，扩展苗绣数字化模拟的应用领域，建立三维苗绣服饰模型，研究其在三维虚拟试衣领域的应用，增加用户的个性化体验。项目的研究扩展了非真实感绘制技术的应用领域，有利于云南苗绣艺术作品的保护和传承，同时有利于推广至其它少数民族非物质文化遗产的数字化模拟。

本项目在前期非真实感绘制(Non-photorealistic Rendering : NPR)相关研究成果的基础上，立足于云南的苗族特色文化，选取国内外享有盛誉的苗绣作品作为研究对象，考察和分析苗绣作品的实际创作过程，对其数字化仿真开展研究。项目建立了苗绣图案数据库，扩展人物、动物、鬼神等图案数据库，通过色彩传输、图象融合、凹凸映射、针迹模拟获得苗绣数字化针迹仿真结果。基于多尺度、双通道等卷积神经网络，对苗绣的针法、绣法进行了数字化模拟，突出了刺绣艺术的线条方向和2.5维立体视觉效果。通过对苗绣艺术作品的数字化合成、显示和优化等问题研究，项目最终模拟了苗绣的图案、色彩、肌理、画面、立体视觉等艺术特征。项目扩展苗绣数字化模拟的应用领域，建立三维苗绣和人物模型，建立三维虚拟试衣系统。此外，项目将苗绣数字化模拟算法扩展到云南蜡染、东巴画，以及粉笔画等艺术风格，获得不同艺术风格的数字化仿真结果，并对小样本的艺术风格图像分类、评价方法进行研究，可应用于艺术风格的搜索、识别等领域。.项目研究进展顺利，完成了预期研究内容和考核指标。发表学术论文21篇（其中SCI收录5篇，EI及源刊收录6篇次）；申请国家发明专利2项，其中1项已授权，获得计算机软件著作权登记3项；获得云南省自然科学奖三等奖1项；在科学出版社出版学术专著2部；开发并实现了苗绣艺术数字化仿真平台、三维虚拟试衣系统；项目组成员入选云南省万人计划青年拔尖人才、云南省中青年学术技术带头人后备人才等；培养青年教师2名，2名博士生、9名硕士生顺利毕业。.项目的研究提升了研究团队研究水平和能力，扩展了非真实感绘制技术的应用领域，有利于云南苗绣艺术作品的保护和传承，同时有利于推广至其它少数民族非物质文化遗产的数字化模拟及应用。