融合人脸轮廓和区域信息改进人脸检测

基于人脸轮廓信息和面部区域信息的互补性，提出了一种新颖的基于融合算法的轮廓一区域人脸检测器：采用一种新的特征提取方法有效地刻画人脸轮廓模式；基于支持向量机分别训练人脸轮廓分类器和面部区域分类器；基于最小错误率Bayes决策规则融合人脸轮廓和面部区域分类器。该文分别在标准头部图像库、BioID人脸图像库(灰度人脸图像库)和彩色人脸图像库上测试了轮廓一区域人脸检测器．大量的实验结果表明了所提出的轮廓一区域人脸检测器通过引入轮廓信息有效地提高了人脸检测算法的精度。

Improving Face Detection through Fusion of Contour and Region Information

Based on the complementarity of face contour and facial region, the paper proposes a novel contour-region face detector. A new feature extraction method is proposed to efficiently depict face contour pattern. Both Face-Contour-Classifier and the Facial-Region-Classifier are trained as Support Vector Machine models. Bayesian decision rule is adopted to fusing both Face-Contour- and Facial-Region- classifiers. The proposed face detector is tested on a standard head image database, BioID gray color face image database and a color face database. Experimental results demonstrate the efficiency of the proposed contour-region face detector.