计算机应用与软件Issue(8)：174-179,6.DOI:10.3969/j.issn.1000-386x.2015.08.041
基于高斯过程回归的多姿态人脸识别
GAUSSIAN PROCESS REGRESSION-BASED FACE RECOGNITION ACROSS POSE
应竞舟 1傅忠谦 1王建宇2
作者信息
- 1. 中国科学技术大学 安徽 合肥 230027
- 2. 上海技术物理研究所 上海200083
- 折叠
摘要
Abstract
On account of the problem that face pose variation will cause serious interference to the recognition result and the fact that currently the majority of face recognition systems only support standard frontal face images, we analyse the mapping relation between non-frontal face profile to frontal face profile based on Gaussian process regression, and propose a method which can handle the face images with pose variation up to 45 degrees to the left or right.This method significantly improves the recognition rate of face recognition system on non-frontal face images.The experiments carried on Mulit-Pie and FERET face database a couple of times show the effectiveness of our method.关键词
人脸识别/姿态校正/计算机视觉/高斯过程回归Key words
Face recognition/Pose normalisation/Computer vision/Gaussian process regression分类
信息技术与安全科学引用本文复制引用
应竞舟,傅忠谦,王建宇..基于高斯过程回归的多姿态人脸识别[J].计算机应用与软件,2015,(8):174-179,6.
