Abstract: LBP algorithm is robust to illumination and can extract?the texture?features of face images. A face recognition algorithm based on LBP and stacked auto encoder is proposed in this paper. At first, uniform LBP features are extracted from different blocks of a face image, which are connected together to serve as the description for the whole face. Then, the LBP feature is input to stacked autoencoders. At last, the trained SAE is used to complete the recognition and classification of face images. The experimental results on Extended Yale B face databases demonstrate that the proposed method is robust to illumination. And it has a better recognition performance compared to traditional algorithms and standard stacked autoencoders.

Key words: face recognition, deep learning, stacked autoencoders, local binary pattern

摘要： LBP算法对光照敏感且能有效地提取图像的纹理结构特征。提出一种基于局部二值模式（Local Binary Pattern，LBP）和栈式自动编码器（Stacked Autoencoders，SAE）的人脸识别算法。用统一模式LBP算子提取分块后的人脸图像的直方图，按顺序连接形成整幅图像的LBP特征，并将其作为栈式自动编码器的输入，完成进一步的特征提取，实现人脸图像的识别与分类。在Extended Yale B等数据库上的实验结果表明，该算法与传统的人脸识别算法和标准的栈式自动编码器相比，对光照变化有更强的鲁棒性，具有更好的识别效果。

关键词: 人脸识别, 深度学习, 栈式自动编码器, 局部二值模式