分层CS-LBP和HOG特征融合的人脸识别研究

Research on Fusion of Layered CS-LBP and HOG for Face Recognition

针对单独的纹理特征只能提取图像的纹理信息而不能得到图像轮廓边缘信息的问题,在人脸识别的特征提取研究中提出了分层CS-LBP和分层HOG特征的融合方法.对图像分别进行多次CS-LBP和HOG特征的提取,得到分层CS-LBP特征提取图像和分层HOG特征提取图像,对分层的特征提取图像再次提取分层CS-LBP和分层HOG特征,并将两种分层特征进行融合,得到更有效的人脸的纹理及边缘轮廓特征.在ORL和GT人脸库上的实验结果验证了所提出的分层特征融合方法的有效性.

Texture feature extracts image texture information only,and is not capable of capturing the edge information,thus a new method of feature extraction is proposed,called layered fusion of CSLBP and HOG feature.Firstly,the CS-LBP and HOG feature are utilized to extract the multi-layered feature from the original image,get different multi-layered feature images;then extract the layered CS-LBP and layered HOG from feature extraction image;finally,mix these two features to represent original image more effectively.The experiment conducted in the ORL and GT standard face image database shows that the fusion method proposed is better than single method,and the rate of recognition has greatly improved.