摘要
针对当前三维头部重建中使用的网格模型不能完整表达头部拓扑结构或者模型过于复杂的问题,提出了一种完整头部网格模型合成方法。该方法首先以Candide-3模型为基础加入后脑勺模型,对面部Candide-3模型的眼睑部分进行重新建模,并加入比例合适的眼球模型,得到一个完整头部网格模型。然后结合人脸特征点信息和局部最小二乘法匹配原理对模型进行调整,并对模型进行纹理贴图合成真实感的虚拟人头部模型。最后,在该模型上用分段线性插值的方法,合成带眼球细节的情感虚拟人表情。实验结果表明,基于该模型合成的虚拟人表情不仅逼真而且情感丰富。
For the problem of head mesh model presently used in three-dimensional reconstruction cannot completely represent the topology of the head or too complex, this paper proposed a full head mesh model synthesis method. This method firstly added the back of the head for the head based on the Candide-3 model, then reconstructed the eyelid of the facial Candide-3 model, and added a proportion appropriate eyeball, thereby a full head mesh model was established. Secondly, it adjustmented the model combining the massage of facial feature points with the matching principle of partial least squares, and synthesized the realistic head model of the virtual human by texture mapping for the head. Finally, it used the method of linear interpolation for the model, and synthesized the expression of the emotional virtual human with details of eyeball. Experimental results show that the synthesized virtual human expressions based on the model look not only realistic but also emotionally rich.
作者
曹宁哲
侯进
马文超
Cao Ningzhe Hou Jin Ma Wenchao(School of Information Science ＆ Technology, Southwest Jiaotong University, Chengdu 611756, China)
出处
《计算机应用研究》
CSCD
北大核心
2016年第12期3839-3842,3867,共5页
Application Research of Computers
基金
成都市科技项目(科技惠民技术研发项目)(2015-HM01-00050-SF)
四川省动漫研究中心2015年度科研项目(DM201504)
国家自然科学基金面上项目(61371165)
关键词
完整网格模型
眼球细节
纹理贴图
表情合成
complete grid model
eye details
texture mapping
expression synthesis
