基于模型的头部运动估计和面部图像合成  

MODEL BASED ESTIMATION OF HEAD MOTION AND FACIAL IMAGE SYNTHESIS

作　　者：尹宝才[1,2] 高文[1,2] 晏洁 宋益波[1,2] 

机构地区：[1]北京工业大学计算机科学与工程系 [2]哈尔滨工业大学计算机科学与工程系

出　　处：《计算机研究与发展》1999年第1期67-71,共5页Journal of Computer Research and Development

基　　金：国家"八六三"高技术计划;国家教委跨世纪优秀人才计划;北京市跨世纪优秀人才工程专项经费;北京市教委科技项目基金

摘　　要：文中讨论一种基于模型的头部运动估计和面部图像合成方法．首先建立了一个基于人脸几何模型的可变形三维面部模型，此模型可根据不同人脸图像特征修正特定人脸模型．为了使特定人脸模型与特定人脸图像相匹配，需根据变形模型修正人脸模型．文中采用自动调整与人机交互相结合的方法实现特定人脸模型匹配．在调整完模型形状之后，应用３个方向的面部图像进行纹理映射生成不同视点方向的面部图像．应用合成面部图像与输入面部图像最佳匹配的方法进行输入面部图像的头部转动方向估计．An approach to the estimation of head motion for the purposes of resynthesizing and coding facial images is presented.A 3 D deformable facial model is introduced for modeling a person's face,which is an interactive generation by modification and matching with the specific human face images.After the model is modified,facial images are generated by using 3 D deformable facial model and multi direction texture mapping.The estimation of head motion is determined by selecting the parameters of head motion such that synthesized facial images in the estimation is the best approximation to the original images.The approach is based on 3 D face model,automatic deformation model.The method is appropriate for model based facial synthetic coding.The estimation and synthesis of local motions,such as mouth,eyes and nose,will depend on motion models of that.Human machine interaction methods need necessary environment and tools.This will be the future works.

关 键 词：面部图像合成 图像编码 面部检测 图像处理 

分 类 号：TN919.8[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]