Abstract: Facing with the problem of face recognition with single training sample per person, the conventional methods will suffer serious performance drop or even fail to work. To solve this problem, this paper proposes an adaptive weighted Local Binary Pattern（LBP） method. It combines both texture feature and topological information with a novel weighted way involving the variance of sub-images. The paper partitions facial images and uses LBP to extract texture feature. It makes use of variance to implement the adaptive weighted fusion for features. The nearest neighbor classifier is adopted for further recognition. Experimental results show a better performance on ORL facial database.

Key words: face recognition, single sample, Local Binary Pattern（LBP）, weightiness

摘要： 在面对单训练样本的人脸识别问题时，传统人脸识别方法识别率会下降很多，有的方法甚至不能使用。针对单样本人脸识别问题，提出了一种自适应加权LBP方法。方法既提取了纹理信息又包含了分块拓扑信息，更重要的是可以把这些特征用合适的权重融合起来。划分图像并用LBP提取纹理信息；利用方差来完成对特征的自适应加权融合；用最近邻分类器识别结果。在ORL人脸数据库上的实验结果表明，该方法可以有效地提高识别率。

关键词: 人脸识别, 单样本, 局部二值模式, 加权