基于视频的漫画人脸动画研究与实现

发布时间：2018-04-18 00:31

本文选题：关键帧 + 人脸动画　；参考：《四川师范大学》2010年硕士论文

【摘要】： 随着手机、互联网等技术的发展和不断完善,数字动漫逐渐形成了当今社会的主要消费主题之一。而人脸动画又是这中间最重要的内容。以人脸为主题的动漫内容的应用也越来越广泛。不但是在我们熟悉的视频会议、网络通信和影视娱乐方面有很好的应用,而且人脸动画越来越多地应用于虚拟教学、个人移动业务等方面。本文设计了一个基于视频的漫画人脸动画生成系统,可以将真实的人脸视频转换成为卡通漫画风格并且时空高度一致的人脸动画。该系统主要分为三大步骤。首先是关键帧的提取。即从视频序列中提取出代表关键动作的关键帧人脸图像。本文在双阈值算法的基础上进行改进,针对人脸部分进行检测并确定出关键帧。使提取的关键帧更注重人脸部分的运动,更准确,更具有代表性。其次是漫画人脸的生成。即对视频图像中的真实人脸进行漫画处理。这部分分为两大步骤实现,即人脸插图画的生成和人脸夸张变形。在人脸插图画生成方面提出一种新的生成人脸插图画算法,使得生成的人脸插图画效果更加自然、更具有艺术感。在人脸夸张变形方面,结合全局变形和局部变形得到了一种更具有喜剧效果和个性化的夸张人脸变形效果。最后是中间帧的插入。即在经过夸张漫画处理的关键帧中间插入中间帧,形成最后的人脸动画。该部分主要将基于域的图像渐变算法应用于中间帧的生成,获得了自然平滑的动画效果。本文阐述了一套基于视频的人脸卡通动画生成系统的设计和实现过程。该系统主要分三个步骤进行实现,包括提取关键帧、生成漫画肖像和生成中间帧。本文论述了实现的相关过程,具有一定的实践价值。
[Abstract]:With the development and improvement of mobile phone, internet and other technologies, digital animation has gradually become one of the main consumption themes in today's society.And face animation is the most important content in the middle.Face as the theme of animation content is also more and more widely used.Not only are we familiar with video conferencing, network communication and video entertainment, but also face animation is more and more used in virtual teaching, personal mobile business and so on.In this paper, a video-based cartoon face animation system is designed, which can transform the real face video into cartoon cartoon style and highly consistent facial animation in time and space.The system is divided into three steps.The first is the key frame extraction.The key frame face image representing the key action is extracted from the video sequence.Based on the double threshold algorithm, this paper detects the face and determines the key frame.The key frame is more accurate and more representative because it pays more attention to the movement of the face part.Secondly, the generation of comic faces.In other words, the real face in the video image is processed by comics.This part is divided into two steps, that is, the generation of face illustration and face exaggeration deformation.In the aspect of human face illustration generation, a new face illustration generation algorithm is proposed, which makes the generated face illustration more natural and more artistic.In the aspect of face exaggeration deformation, combining global deformation and local deformation, a more comic and personalized hyperbolic face deformation effect is obtained.Finally, the middle frame is inserted.The final face animation is formed by inserting the middle frame between the key frames processed by hyperbole comics.In this part, the image gradient algorithm based on domain is applied to the generation of middle frame, and the natural smooth animation effect is obtained.This paper describes the design and implementation of a video-based face cartoon animation generation system.The system is mainly implemented in three steps, including extracting key frames, generating cartoon portraits and generating intermediate frames.This paper discusses the implementation of the relevant process, has a certain practical value.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：TP391.41

【参考文献】