面向人脸光照合成的光照参数精确估计算法

刘剑毅; 何苗; 马瑶; 刘跃虎

doi:10.7652/xjtuxb201306004

您当前的位置：

首页 >

文章列表页 >

面向人脸光照合成的光照参数精确估计算法

更新时间：2025-05-01

- 面向人脸光照合成的光照参数精确估计算法
- An Accurate Algorithm to Estiamte Illumination Parameter for Facial Relighting Synthesis
- 西安交通大学学报 2013年47卷第6期页码：18-24
- 作者机构：
  
  西安交通大学人工智能与机器人研究所,西安,710049
- 作者简介：
- 基金信息：
  
  : 2012-10-19。作者简介: 刘剑毅(1975—),男,博士,讲师。基金项目: 国家自然科学基金资助项目(60903122,91120009);教育部高等学校博士学科点专项科研基金资助项目(20090201110029)
- DOI：10.7652/xjtuxb201306004
  中图分类号： TN914.42
- 网络首发：2013-06-10，
  
  纸质出版：2013
- 稿件说明：
移动端阅览
刘剑毅, 何苗, 马瑶, 等. 面向人脸光照合成的光照参数精确估计算法[J]. 西安交通大学学报, 2013,47(6):18-24.

An Accurate Algorithm to Estiamte Illumination Parameter for Facial Relighting Synthesis[J]. 2013, 47(6): 18-24.
刘剑毅, 何苗, 马瑶, 等. 面向人脸光照合成的光照参数精确估计算法[J]. 西安交通大学学报, 2013,47(6):18-24. DOI： 10.7652/xjtuxb201306004.

An Accurate Algorithm to Estiamte Illumination Parameter for Facial Relighting Synthesis[J]. 2013, 47(6): 18-24. DOI： 10.7652/xjtuxb201306004.

摘要

为提高人脸图像光照合成的精度

提出了一种基于商图像的光照参数估计新算法。首先分析并指出传统商图像算法中存在恒定反射系数假设难以严格成立这一缺陷

它给光照参数估计引入了固有误差; 然后在参数化光照子空间框架下设计了一个改进的目标函数

用商图像本身去替代原算法中的反射系数比

从而弱化了假设条件的限制

允许人脸各像素点拥有独立的反射系数; 设计了一个迭代算法

通过线性系统来求解目标函数

能够实现对光照参数的高精度估计。在OFD人脸库和AI&R人脸库上的实验结果表明

所提光照参数估计算法具有良好的收敛性

光照参数估计精度的迹商指标可以获得25.49～43.09 dB的提升

光照合成图像中的峰值信噪比可以获得4.51～9.24 dB的提升。

Abstract

A novel algorithm to estimate illumination parameter based on quotient image(QI)is proposed to improve the accuracy of facial image illumination synthesis. Analysis indicates that the hypothesis of uniform albedos in traditional QI algorithm can not be strictly hold

and that inherent error in illumination parameter estimation will be incurred. Then

an improved objective function is derived in the parametric illumination subspace framework

and the albedo ratios is replaced by the quotient image itself

so that the impacts of the hypothesis is weakened

and independent albedo per facial pixel can be permitted. An iterative algorithm is designed to solve the objective function through linear system

and illumination parameter can be estimated at high accuracy. Experiments on the oriental face database(OFD)and the AI&R face database show that the proposed algorithm has good convergence ability

and that the trace ratio for illumination estimation and the peak-signal-to-noise-ratio in illumination synthesis images are improved by 25.49～43.09 dB and 4.51～9.24 dB

respectively.

关键词

Keywords

references

SHIM Hyunjung, LUO Jiebo, CHEN Tsuhan. A subspace model-based approach to face relighting under unknown lighting and poses [J]. IEEE Transactions on Image Processing, 2008, 17(8): 1331-1341.

SHASHUA A, TAMMY R R. The quotient image: class-based re-rendering and recognition with varying illuminations [J]. IEEE Transactions on PAMI, 2001, 23(2): 129-139.

熊鹏飞, 刘昌平, 黄磊. 一种人脸标准光照图像的线性重构方法 [J]. 模式识别与人工智能, 2012, 25(4): 656-663.

XIONG Peng Fei, LIU Chang Ping, HUANG Lei. A linear reconstruction method for face images under normal illumination [J]. Pattern Recognition and Artificial Intelligence, 2012, 25(4): 656-663.

李粉兰, 段海峰, 郝建国, 等. 人脸识别中光照补偿问题的实验研究 [J]. 工程图学学报, 2009(3): 113-120.

LI Fenlan, DUAN Haifeng, HAO Jianguo, et al. The experimental research of illumination compensation problem in face recognition [J]. Journal of Engineering Graphics, 2009(3): 113-120.

CHEN Terrence, YIN Wotao, ZHOU Xiang Sean, et al. Total variation models for variable lighting face recognition [J]. IEEE Transactions on PAMI, 2006, 28(9): 1519-1524.

王晶, 苏光大. 基于方向滤波和自商图像的人脸光照补偿 [J]. 光子学报, 2010, 39(9): 1641-1644.

WANG Jing, SU Guangda. Illumination compensation based on direction filter and self quotient imag [J]. Acta Photonica Sinica, 2010, 39(9): 1641-1644.

WANG Haitao, LI Stan Ziqing, WANG Yangsheng, et al. Self quotient image for face recognition [C]∥Proceedings of the International Conference on Image Processing. Piscataway, NJ, USA: IEEE, 2004: 1397-1400.

WANG Haitao, LI Stan Ziqing, WANG Yangsheng. Generalized quotient image [C]∥Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2004: 498-505.

AN Gaoyun, WU Jiying, RUAN Qiuqi. An illumination normalization model for face recognition under varied lighting conditions [J]. Pattern Recognition Letters, 2010, 31(9): 1056-1067.

何晓光, 田捷, 毋立芳, 等. 基于形态学商图像的光照归一化算法 [J]. 软件学报, 2007, 18(9): 2318-2325.

HE Xiaoguang, TIAN Jie, WU Lifang, et al. Illumination normalization with morphological quotient image [J]. Journal of Software, 2007, 18(9): 2318-2325.

LIU Jianyi, ZHENG Nanning, XIONG Lei, et al. Illumination transition image: parameter-based illumination estimation and re-rendering [C]∥Proceedings of the International Conference on Pattern Recognition. Piscataway, NJ, USA: IEEE, 2008: 1-4.

PIZER S M, AMBURN E P, AUSTIN J D, et al. Adaptive histogram equalization and its variations [J]. Computer Vision Graphics and Image Processing, 1987, 39(3): 355-368.

BEIER T, NEELY S. Feature-based image metamorphosis [J]. Computer Graphics, 1992, 26(2): 35-42.

西安交通大学. OFD东方人脸数据库. [DB/OL].(2007-04-08)[2012-08-19]. http:∥www.aiar.xjtu.edu.cn∥dfrlsjk5.html.

FU Raymondyun, ZHENG Nanning. M-Face: an appearance-based photorealistic model for multiple facial attributes rendering [J]. IEEE Transactions on Circuits and Systems for Video Technology, 2006, 16(7): 830-842.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

电动车辆间接热系统快速除湿响应迟滞机理及优化控制策略

采用试验设计方法的第三代CO2气体箔轴承箔片几何参数敏感性分析

离子液体改性燃料的电流体动力学效应研究

采用机器学习的大规模并行程序I/O性能建模与预测方法

冷库水平空气幕热湿阻隔性能数值模拟研究