节点文献

3D虚拟人脸的自动生成及情感表达

3-D facial virtual character auto-construction and emotion expressing

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 邬正义陆建明施健郭春霞谈正

【Author】 Wu Zhengyi1+, Lu Jianming1, Shi Jian1 , Guo Chunxia2 ,Tan Zheng1,2 1(Digital Technique Research Group, Changshu Institute of Technology ,Jiangsu 215500) 2 (Research-Institute of Information Engineering, Xi’an Jiaotong University, Xi’an 710049)

【机构】 常熟理工学院数字技术综合研究所西安交通大学电信学院信研所

【摘要】 首先从正面图像中自动检测并定位人脸面部特征点,并根据正面人脸属性估计特征点的深度信息,再根据特征点坐标进行体形变处理,实现网格个性化,最后通过对人脸图像进行的柱面纹理映射,完成三维虚拟角色自动生成并实现表情可控表达。本方法在一定姿态变化范围内可获得理想的视觉效果,同时具有快速、方便、无需人工干预和硬件设备要求低等优点。

【Abstract】 3-D virtual character auto-construction has been widely used in virtual reality, intelligent human-computer interface, face recognition, facial motion image design and so on. Constructing 3-D facial model of specified face according to facial multi-visual angle image based on generic facial 3-D grid has been researched and has received some applications. However, this method requires much to the hardware and needs human-computer interaction, which has limited its application fields. Considering the statistical relationship between the facial depth and obverse information, and the fact that depth information has few influence to visual effect in certain pose range, this paper proposes a 3-D facial virtual character construction method based on single obverse face image. Firstly the facial features are detected based on the relation between 2-D face image and general 3-D grid model, then the depth of the feature point are estimated according to the extracted frontal properties and the generic model is individualized by volume morphing with a novel Radial Basis Function-based interpolation, finally the 3-D virtual character is constructed through the texture cylindrical mapping. The controllable affective expression of the virtual character can be realized by some specified point in the grid model. This system can construct corresponding 3-D facial virtual character automatically based on single obverse face image, which can realize affective expression of controllable pose. The virtual character can rotate up, down, left and right in the range of ±30o, which has emotions such as happiness, joy, grief, anger and so on. This system is fast, convenient and with low requirement, which make it a potential solution in virtual reality and intelligent human-computer interface. The virtual character construction process and the controllable affective expression can demonstrate by motion pictures.

【基金】 江苏省高校自然科学研究计划项目(05KJD510004)
  • 【会议录名称】 第二届和谐人机环境联合学术会议(HHME2006)——第15届中国多媒体学术会议(NCMT’06)论文集
  • 【会议名称】第二届和谐人机环境联合学术会议(HHME2006)——第15届中国多媒体学术会议(NCMT’06)
  • 【会议时间】2006-10
  • 【会议地点】中国浙江杭州
  • 【分类号】TP391.41
  • 【主办单位】清华大学计算机科学与技术系、浙江大学计算机科学与技术学院
节点文献中: 

本文链接的文献网络图示:

本文的引文网络