基于层次化显著性与年龄演化特征的逆向人脸美化方法

项目名称： 基于层次化显著性与年龄演化特征的逆向人脸美化方法

项目编号： No.61472245

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 马利庄

作者单位： 上海交通大学

项目金额： 84万元

中文摘要： 本项目围绕基于层次化显著性与年龄演化特征的逆向人脸美化中的若干关键技术展开深入研究。在传统显著度计算算法的基础上，结合颜色、纹理和几何等特征，研究利用层次化原理结合特征区域来形成可视媒体大数据中人脸图像与视频的显著度图；计算影响年龄的人脸显著性区域，提取与年龄相关的显著性特征，建立有效人脸年龄特征模型；研究适用于不同人种或种族的人脸生长模型，建立用于表征人脸年龄演化影响的计算方法，对人脸年龄演化进行高效建模；将人脸抽象成特征点的的组合，通过年龄演化模型对特征点之间的关系进行标准化，实现逆年龄演化的人脸美化，有效解决人脸美化中的严重失真和美化程度不一致等问题，使得基于年龄的人脸层次化显著性特征美化技术更加自然合理；研究并实现逆年龄演化的人脸美化关键算法及原型系统。本项研究可望推动数字娱乐、网络社区等可视媒体大数据领域中人脸美化理论及应用技术的发展，具有重大理论研究和实际应用意义。

中文关键词： 计算机图形与图像；逆向人脸美化；年龄演化模型；层次化显著性特征

英文摘要： This project studies deeply on some key techniques about the reverse-age face beautification based on hierarchical saliency and age evolution characteristics. Based on the traditional algorithm to compute significance, and combined with color, texture and geometric features, we study the saliency map of face images and videos in big data of visual media by using the hierarchical principle combined with the characteristic area; calculate the face significant region related with age, extract the significant characteristics of age related, establish effective model for age feature of face; research on face growth model, find the algorithm for representing the face age evolution, and simulate efficiently for face age evolution; abstract the face as the combination of feature point, standardize the relationship between the feature points according to the age evolution the model, realize the face beautification of reverse-age, solve effectively the problem of serious distortion and age inconsistency in face beautifying , so that the face beautification techniques based on hierarchical saliency and characteristics of face age are more natural and reasonable; study and implement the face beautification algorithms and prototype system of reverse-age evolution. This project is expected to promote the development of facial beautification techniques and related applications in digital entertainment, online communities and the other big data of visual media. It has great significance in both theory and applications.

英文关键词： Computer Graphics and Image Processing;Reverse-age Face Beautification;Age Evolution Modeling;Hierarchical Saliency