权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Face Analysis and Synthesis Model Expressing Individual Characteristics and Aging Features with High Reality

高真实度表达个体特征和衰老特征的人脸分析与合成模型

基本信息

批准号：
22300070
负责人：
MORISHIMA Shigeo
金额：
$ 11.81万
依托单位：
Waseda University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
2010
资助国家：
日本
起止时间：
2010 至 2012
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-22300070/
关键词：
ディジタルヒューマンモデル顔の分析・合成解剖学的顔モデル年齢・性別変換顔認証３次元顔形状推定コンピュータビジョン画像情報処理顔形状3次元復元年齢変化シミュレーション表情合成モデルデジタルヒューマンモデル

项目摘要

In the several properties of human face, we focused on personal characteristics, age, emotion and health condition, and then a parametric face model has been realized with high reality and accuracy. First of all, we produced an individual face model with variation of muscles location and fat layer structure on face by anatomical analysis of real face by collaboration with Prof. Shimada. Next, we found the statistical rules in a human face structure and constructed a generic 3D face model by analysis of huge face database including 3D range scan data and frontal snap shots. At first, we proposed the algorithm to extract feature points stably and robustly on frontal face image by linear predictors. Then we proposed new method to extract individual 3D feature of face accurately and constructed a real-time and automatic 3D face reconstruction system by camera captured photo. Also we found the wrinkles are very important to produce an aged face, and then we constructed an interactive aged face simulator by controlling shape of face, textures and wrinkles. Moreover, based on this proposed face model generation platform, we contributed to the research area of face image processing strongly by proposing a face authentication system robust to aging effect, driver’s drowsiness estimation system, face image synthesis system with wrinkles control, lip synchronization system and face recognition system between natural and artificial smile.

在人脸的几个属性中，我们重点关注了个人特征、年龄、情绪和健康状况，从而实现了具有较高真实感和准确性的参数化人脸模型。首先，我们与Shimada教授合作，通过对真实面部的解剖分析，制作了一个具有面部肌肉位置和脂肪层结构变化的个体面部模型。其次，通过分析包括三维距离扫描数据和正面快照在内的庞大人脸数据库，找到人脸结构的统计规律，构建通用的三维人脸模型。首先，提出了利用线性预测器稳定鲁棒地提取正面人脸图像特征点的算法。在此基础上，提出了一种准确提取人脸个体三维特征的新方法，构建了一个实时、自动的三维人脸重建系统。同时，我们发现皱纹对于老年脸的生成是非常重要的，然后我们通过控制面部形状、纹理和皱纹构建了一个交互式的老年脸模拟器。此外，基于该人脸模型生成平台，我们提出了抗衰老人脸认证系统、驾驶员睡意估计系统、皱纹控制人脸图像合成系统、嘴唇同步系统和自然与人工微笑人脸识别系统，为人脸图像处理研究领域做出了重要贡献。