基于特征提取与认证的彩色图像人脸检测

摘要

The method of basing on the whole character face image mainly has Eigenface,fisherface,spectroface and NN etc.But the single method＇s recognition accuracy and recognition velocity remain to be enhanced.It＇s some distance to the practical application;therefore it needs to continuously improve the method.After deeply studying the method of spectroface and fisherface,the thesis extracts the advantages of the two methods,present a new face recognition method based on the spectroface and fisherface.Spectroface representation combines the wavelet transform and the fourier transform.By decomposing a face image using wavelet transform,the low-frequency face image is less sensitive to the facial expression variations.By decomposing the low-frequency face image using fourier transform,the face image with lower-dimensional space is obtained.But the dimension of the spectroface is still high,so extracting the character of Fisherface from the spectroface in order to reduce the dimension and enhance the efficiency.Structure generated using human facial feature extraction gray eyes,a mouth of the color characteristics of segmented mouth,and then under the eyes and mouth features constitute a triangle template,accurately locate the face position in the image.Experimental results show that the combination of skin color and facial features algorithm is able to face a more rapid and accurate positioning,and the results are more reliable.%在深入的对频谱脸法和Fisherface方法进行研究后,综合这两种方法的优点,提出了一种基于频谱脸和Fisher-face的人脸识别新方法。频谱脸方法主要是采用二维小波变换和傅立叶变换。因为人脸图像的低频部分对人脸的表情变化是不敏感的,所以对人脸图像使用二维小波变换,提取人脸图像的低频部分。对人脸图像的低频部分使用傅立叶变换,从而获得原人像的一个低维空间的表达。但是频谱脸特征维数仍然较高,所以在频谱脸法的基础上继续提取人脸频谱图像的Fisherface特征,降低特征的维数,提高识别效率。利用人脸面部构造产生的灰度特性提取眼睛,利用嘴唇的色度特征分割出嘴巴,进而根据眼睛和嘴巴构成三角形模板的特性,精确定位人脸在图像中的位置。实验结果表明,这种结合肤色和面部特征的算法,能够对人脸进行较快速、准确的定位,而且结果比较稳定可靠。