基于认知机理的三维人脸建模及应用研究

Perception based 3D face modeling and its applications

人类天生具备从单张平面照片进行三维信息恢复的能力,模拟人类的这一认知过程,研究基于单张平面图像的人脸三维建模技术,是当前认知计算中的一个重要问题。以统计学为工具,在建立人脸统计模型的基础上提出一个两步人脸形状建模方案,利用人脸结构先验知识对建模过程进行约束,能够生成真实感强的三维人脸模型。将三维人脸模型应用到光照和姿态估计问题上,取得了较好的效果。

Human beings are born with a capacity of restoring the 3D shape from a single planar image. It＇s an essential research topic in perceptual computing to develop novel approaches for modeling 3D human face from a 2D image via simulating this perceptual procedure. Using the statistic theory, this paper proposes a two-step face modeling scheme based on a facial shape statistic model, which serves as the prior knowledge to constrain the following modeling steps. Realistic 3D faces could he generated with this method. Good results have achieved by applying 3D faces to the problem of illumination and pose estimation.