Abstract: Local Binary Pattern（LBP） has been used with considerable success in face recognition field, while it hasn’t considered the contrast between pixels so that some important texture features have been abandoned. This paper proposes to calculate dynamic thresholds of the contrast range by coarse grain algorithm, and extracts the face feature with the thresholds, meanwhile the code method is similar to Local Multi-layer Contrast Pattern（LMCP）. Moreover, Local Nonlinear Multi-layer Contrast Pattern（LNMCP） can get the face features, which contained the information of illumination various. The experiments show that the proposed method has a high recognition rate, the results on outdoor database can be enhanced by 2.91% especially. It proves that the method will be adaptive under uncontrollable varying illumination.

Key words: Local Nonlinear Multi-layer Contrast Pattern（LNMCP）, Local Binary Pattern（LBP）, coarse grain algorithm, dynamic threshold

摘要： 局部二值模式（LBP）在人脸识别领域取得了显著成效，但由于没有考虑像素之间的对比度，导致部分重要纹理特征被丢弃。采用粗粒度分区算法求取各人脸分块对比度值域区间的动态阈值点，并借鉴LMCP算法思想利用上述阈值点分别对各人脸分块编码，得到结合光照变化信息的局部非线性多层对比度特征LNMCP作为人脸特征。实验结果表明，该方法识别率高，特别是在户外人脸库上提高了2.91%，方法在不可控光照变化环境下具有极强的自适应性。

关键词: 局部非线性多层对比格局, 局部二值模式, 粗粒度算法, 动态阈值