烙画艺术模拟及其数字合成技术研究

。 Non-Photorealistic Rendering (NPR) combines artist painting rule and some scientist methods. This technique can express and transfer information besides photorealistic computer graphics. NPR covers lots of fields, such as simulating nature scene or artist's painting work, enhancing image's information, creating animation, expressing art of date. In addition, the technology of NPR plays an increasingly important role in medical science, architectonic and education. The appearance of NPR research fields enriches the content of graphics. Recently, some western art style such as oil, watercolor pencil drawing etc have been simulated. Because of the demands of real time and diversity of business circle, some researchers also simulated art such as Chinese painting, Chinese ink etc. However, there is still lack of research on Chinese national pyrography art style. Based on our former research work of NPR, we will render and simulate pyrography stylized art. During the development of the proposal, we will analyse the creative process of real pyrography work, and reveal several specialties of this artistic effect, such as line simulating, prospect and color control etc. Firstly, we will put forward the corresponding digital synthesis algorithm about pyrography art in smooth line, enjoyable frame, stereoscopic impression etc. Secondly, we will extend the algorithm to video scene. The most important thing about artistic video synthesis is to preserve smoothness in transition. By calculating a weighted similarity among past and future frames, we choose adapted loop sequence and transtion among frames will happen in a proper order. Based on the technique of frame segmentation and frame compensation, the proposal will also eliminate visual discontinuities between neighbor frames. In addition, based on CUDA we will research accelerated and optimized algorithms about pyrography style and video digital synthesis system. The related research findings will provide a new NPR artistic field of pyrography style. Our mehtod will provide a generic mechanism for research different NPR artistic effects in video synthesis. The project is expected to publish 10-15 papers in national and international journals and conference, and develop realtime artistic synthesis system of pyrography effect and video scene.

对不同艺术风格的数字化模拟是非真实感绘制技术的主要目标之一。近年来非真实感绘制的研究对象不断扩展，研究成果广泛应用于虚拟现实、网络传输、影视制作等方面。国内外很多研究者已对中国特有的书法、水墨画等艺术作品进行了数字仿真，但尚缺乏对具有民族和地域特色的烙画艺术作品的模拟研究。本项目选取烙画艺术风格作为研究对象，分析和考察烙画艺术的实际创作过程，揭示其线条、色彩等方面在创作过程中的规律和特点，研究基于图像的烙画艺术风格数字合成关键技术，重点针对烙画作品中流畅细腻的线条、写意的画面、凹凸立体感、创作材质等创作技法提出相应的模拟算法；其次将算法扩展到视频场景，研究视频帧分割、视频帧补偿等关键问题，同时基于CUDA建立烙画及其视频的实时绘制系统。相关研究成果是对非真实感绘制技术的有益补充，对其它民族特色的艺术风格数字化模拟具有借鉴作用，为不同艺术效果在视频合成中的研究提供了一种通用的实现机制。

项目选取烙画艺术风格作为研究对象，分析和考察烙画艺术作品的实际创作过程和艺术特征，揭示其线条、色彩等方面在创作过程中的规律和特点，对烙画艺术风格作品进行数字化仿真。项目首先构建烙画前景和材质数据库，基于纹元的语义特征和纹理合成技术，扩展烙画背景材质图像；其次，针对烙画前景图像的艺术特征进行数字化模拟，实现了抽象写意的画面、流畅细腻的线条以及画面的凹凸立体感，获得前景图像艺术特征；接着，提出采用光照明模型对艺术画面前景图像与材质背景图像进行快速融合，获得具有烙画艺术风格的结果图像，项目通过二维图像变形算法，产生葫芦烙画艺术风格图像。此外，项目提出基于置信度光流场的视频帧图像分割算法、基于内容感知的视频帧补偿算法，将烙画艺术风格效果扩展至视频场景，为视频场景注入了艺术效果展示能力。项目通过CUDA等对绘制算法加速，同时采用视觉感知、信息熵等方法对模拟的艺术风格作品进行评价，建立烙画仿真评价标准。. 项目开展顺利，在国际国内会议和期刊上发表学术论文21篇，其中SCI收录5篇，EI收录9篇次，国内外核心刊物13篇；申请国家发明专利5项，其中已授权3项，申请并获准计算机软件著作权10项；获得云南省自然科学奖三等奖1项（排名第二、第三）；出版学术专著2部；项目组成员2人晋升为教授，2名博士生获得博士学位，4名硕士生获得硕士学位，培养学术骨干2名。项目研究完成了任务书中所拟定的各项研究内容、预期目标。相关研究成果是对非真实感绘制技术的有益补充，对其它具有民族特色的艺术风格数字化模拟具有借鉴作用，对弘扬和保护中国民族文化，促进影视作品、虚拟现实、网络数据传播等领域的发展具有积极作用。