中文摘要：

子模式主成分分析（SpPCA）算法忽略了人脸不同分块应该具有不同的重要性。为了解决此问题．提出一种自适应加权SpPCA单样本人脸识别算法，对人脸图像的不同分块自适应地计算其权重。算法对人脸进行分块，按照SpPCA算法将各个分块投影到特征脸的基坐标上，并以每个模块LBP编码的纹理图像信息熵来表征该模块的权值；将模块的权重赋予该模块的特征脸投影，并得到最终分类结果。实验在YaleB和扩展YaleB人脸数据集上进行测试。实验表明，该算法得到了较好的识别结果，有效地弥补了SpPCA算法的不足。

英文摘要：

The sensitivity to illumination changes is one of the most important issues for the evaluation of face recognition sys- tems. This paper proposed a new approach to recognize face images under variation of lighting conditions when only one sample image per person was available. This approach represented a face image as an array of sub-pattern PCA （ spPCA ） extracted from a partitioned face image containing information of local regions. In order to adjust the contribution of each local region of a face in terms of the richness of identity information, it utilized LBP based information entropy weighting technique to assign proper weights to SpPCA features. The experimental results use Yale B and extended Yale B databases to demonstrate that ro- bustly recognizing face images under different lighting conditions.