当前位置:主页 > 文艺论文 > 广告艺术论文 >

人脸器官拼接融合及其在人脸动画中的应用

发布时间:2018-02-12 03:06

  本文关键词: 人脸动画 MPEG-4人脸动画标准 Mesh-Warping算法 语音识别 图像拼接 图像融合 肤色模型 特征点定位 出处:《中国科学技术大学》2009年硕士论文 论文类型:学位论文


【摘要】: 人脸动画是广泛应用于计算机动画行业、游戏行业、远程会议、代理和化身等许多领域,是近几年来国内外研究的热点。其中,基于单幅图像的人脸动画的研究近年来引起了许多研究者的关注。基于单幅图像的人脸动画,对给定的一幅图像进行几何位置上的变换,能产生一些有趣的视觉效果,在影视娱乐、广告设计中常用来产生特技。 基于单幅图像的人脸动画由于是在真实人脸图片的基础上做图像变形而产生,所以具有较高的真实感。本文提出并实现的一种基于单幅图像的人脸动画生成系统,在MPEG-4人脸动画标准的基础上,采用了肤色模型等人脸特征点定位技术和Mesh-Warping的图像变形方法生成人脸动画,在口腔/眼球的处理上采用了图像拼接和融合方法,使得生成的人脸动画在保证实时性的同时获得了较为逼真的动画效果,本文主要内容包括: 1)调研并总结了人脸动画中的MPEG-4人脸动画标准、人脸特征点定位、图像变形和图像拼接融合技术,分析对比了这些技术的优缺点。 2)实现了一个具有较高真实感和实时性的基于单张人脸图像的人脸动画生成系统。将MPEG-4中的人脸动画标准简化后应用到基于单张人脸图像的2D人脸动画生成中。可对人脸进行自动/手动定位。选用了Mesh-Warping算法对图像进行变形,既保持了较高真实感,又能满足系统的实时性要求。 本文的主要创新有: 1)在Mesh-Warping变形算法中,根据人脸器官运动的特殊性,在构造变形样条时,提出了整体样条和局部样条结合,横向样条和纵向样条分离的方法。在保持图像变形效果的同时,尽可能的减少了运算时间。 2)提出了在极坐标上基于样条插值的人脸器官拼接算法。该算法可以较快的完成人脸口腔/眼睛器官样本图像与人脸图像的拼接和融合,使动画中出现张嘴效果时,能看较为清楚较为自然地看到舌头和牙齿等原图没有的器官。并且算法具有灵活性,通过选用不同的样条,可以在变形效果和计算时间之间寻找符合用户要求的折中点。 3)通过该拼接融合算法,能将样本图像上的眼睛移植到原图像上,以满足某些特殊的动画要求,并且移植后的眼睛在眨眼时能保持形状不变,还可以做出像对眼这样的眼球运动的动画。
[Abstract]:Face animation is widely used in computer animation industry, game industry, teleconferencing, agent, avatar and many other fields. In recent years, the research of face animation based on single image has attracted the attention of many researchers. Based on single image, the geometric position of a given image can be transformed, which can produce some interesting visual effects. In film and television entertainment, advertising design is often used to produce special effects. The face animation based on single image has a high sense of reality because it is based on the deformation of real face image. This paper proposes and implements a face animation generation system based on single image. On the basis of MPEG-4 face animation standard, face animation is generated by skin color model and Mesh-Warping image deformation method, and image stitching and fusion are used in oral / eyeball processing. So that the generated face animation not only guarantees real-time, but also gets a more realistic animation effect. The main contents of this paper are as follows:. 1) the MPEG-4 face animation standard, face feature point location, image deformation and image mosaic fusion technology in face animation are investigated and summarized, and the advantages and disadvantages of these techniques are analyzed and compared. 2) A face animation generation system based on single face image with high realism and real time is implemented, and the face animation standard in MPEG-4 is simplified and applied to 2D face animation generation based on single face image. The face can be automatically / manually located. The Mesh-Warping algorithm is used to deform the image. It not only keeps a high sense of reality, but also meets the real-time requirements of the system. The main innovations of this paper are:. 1) in the Mesh-Warping deformation algorithm, according to the particularity of face organ motion, a new method of combining global spline and local spline, separating transverse spline and longitudinal spline, is proposed in the construction of deformed spline, which keeps the effect of image deformation at the same time. Minimizes computational time as much as possible. 2) an algorithm of facial organ splicing based on spline interpolation in polar coordinates is proposed. This algorithm can quickly complete the stitching and fusion of human face oral / eye organ sample image and face image, so that when the mouth opening effect appears in the animation, We can see clearly and naturally the organs in the original images such as tongue and teeth, and the algorithm is flexible. By selecting different splines, we can find a compromise point between the deformation effect and the computing time. 3) through the fusion algorithm, the eyes on the sample image can be transplanted to the original image to meet some special animation requirements, and the transplanted eyes can keep the same shape in the blink. You can also animate eye movements like eye-to-eye movements.
【学位授予单位】:中国科学技术大学
【学位级别】:硕士
【学位授予年份】:2009
【分类号】:TP391.41

【相似文献】

相关硕士学位论文 前1条

1 倪奎;人脸器官拼接融合及其在人脸动画中的应用[D];中国科学技术大学;2009年



本文编号:1504645

资料下载
论文发表

本文链接:https://www.wllwen.com/wenyilunwen/guanggaoshejilunwen/1504645.html


Copyright(c)文论论文网All Rights Reserved | 网站地图 |

版权申明:资料由用户91738***提供,本站仅收录摘要或目录,作者需要删除请E-mail邮箱bigeng88@qq.com