人脸的表情与年龄变换和非完整信息的重构技术(上)  

Facial Expression Transformation,Aging and Invisible View Reconstruction (1)

作　　者：郑南宁[1] 付昀[1] 张婷[1] 卓峰 

机构地区：[1]西安交通大学人工智能与机器人研究所,陕西西安710049

出　　处：《电子学报》2003年第z1期1955-1962,共8页Acta Electronica Sinica

基　　金：国家自然科学基金创新研究群体 (No .60 0 2 4 30 1 );国家自然科学基金(No.60 2 0 50 0 1 );2 0 0 2 2 0 0 3年度河南省重大科技基金 (No.0 2 2 2 0 2 0 4 0 0 )

摘　　要：人脸表情变换、年龄变换和非完整信息重构是计算机视觉与图像处理领域的重要研究内容 .该文研究了基于线性统计学模型的真实感人像变换和处理技术 .应用对人像形状与纹理的矢量化表达 ,并结合Lambertian表面比率图理论 ,提出了基于稠密特征对应的人脸复杂表情的合成与克隆算法 ;根据人像平均化最美的生理学理论 ,提出了基于平均脸的人像年轻化变换算法 ,以及基于衰老纹理映射和差异性理论的人像衰老化变换算法 .以MPI高加索人像和AIAR东方人像库数据为测试样本的实验结果 ,验证了人像表情合成、衰老化与年轻化算法的有效性 .Facial expression synthesis,facial aging and the reconstruction of the invisible facial views are important research contents in the computer vision and computer graphics fields.This paper addresses the statistical model-based techniques of facial image transformation and processing.With the pixel-wise correspondence computing,the facial image is represented as coupled shape-texture vectors.Based on the quotient image of Lambertian surface classes,an algorithm for realistic expression synthesis and clone is first proposed.Then,the new technique for face juvenility transformation from average face,according to the related hypotheses for face attractiveness in Psychology,is presented.Finally,an ageing transformation algorithm is proposed from merging the age template mapping and the distinctiveness caricaturing methods.With the examples of MPI Caucasian Face and AIAR Asian Face databases,realistic and natural experimental results validate the approaches for the facial expression synthesis and facial age transformation.

关 键 词：人像变换 人像处理 表情变换 年龄变换 视点变换 

分 类 号：TN911.73[电子电信—通信与信息系统]