摘要
因易受到姿态等因素的影响,传统二维人脸识别的结果往往难以令人满意。针对不同姿态下的三维人脸深度图提出一种新的图像处理方法:首先利用面部特征点定位和偏转角度的计算将人脸深度图旋转到"正中面"位置;然后根据人脸对称性和ICP相关理论,对旋转后的深度图进行插值。通过该方法处理后的三维人脸深度图与原始数据相比,在人脸识别实验中具有较好的识别效果和鲁棒性。
Traditional 2D human face recgoniton are always less than satisfactory because of the varied poses. This paper presents a new image analysis method for 3D human face depth maps in different poses. The facial depth maps were rotated to standardized front and neural place according to facial features locating and rotating angles calculation, the facial symmetry and ICP algorithm were also employed to restore depth maps. Recognition experiments with the adjusted 3D face depth maps outperform those with the raw depth maps.
出处
《电子测量与仪器学报》
CSCD
2011年第10期870-878,共9页
Journal of Electronic Measurement and Instrumentation
基金
教育部博士点基金项目(20100111110005)
安徽省2010高校省级自然科学研究重点项目(KJ2010A193)
关键词
图像处理
人脸识别
三维成像系统
深度数据
对称性
image processing
face recognition
3D facial imaging system
depth data
symmetry
