摘要
本文根据证件照片中脸部各个对象的不同特征,分别采用不同的定位方法定位各个特征点。眼睛和嘴巴部位采用SUSAN算子检测眼角和嘴角,鼻子采用水平和垂直积分投影定位鼻孔,眉毛部位先用Canny算子检测边缘,然后对边缘图像进行积分投影定位眉毛的左右边缘点。特征点定位后进行各对象的特征提取,最后利用选定的参数对各个特征进行归一化。实验结果表明,针对各个特征所采用的方法都具有很好的效果,平均定位准确率达到95%以上。
Base on the different facial characters of ID photo, the paper proposes several methods to locate leature points for extracting the facial feature exactly. SUSAN operator is used to locate eye and mouth comer. Nose is located by gray level projection. Canny operator is adopted to detect the edge of eyebrow, and then gray level projection is applied to locate eyebrow edge points. After locating every feature points, different facial parameters are calculated. At last, the parameters are normalized by selected parameter. The experiment result shows that the method＇s accuracy can be great than 95% averagely.
出处
《微计算机信息》
北大核心
2006年第07S期200-202,共3页
Control & Automation
基金
上海市教委曙光项目(04CX72)
关键词
图像分割
特征提取
SUSAN算子
积分投影
Image segmentation, Feature extraction, SUSAN operator, Gray level projection
