摘要
局部二值模式(LBP)是一种对灰度图像的纹理进行描述的方式。然而经典的LBP算法存在数据使用不充分的问题，因此提出一种分层的LBP算子（HLBP）。HLBP选取多层阈值对图像进行LBP序列化，然后将序列化后的LBP直方图进行联合。并在人脸识别中应用改进后的HLBP算子提取特征。基于YALE和ORL人脸数据库的实验结果表明，HBLP算法比经典的LBP算法的识别率高。
Abstract
LBP is defined as a texture description method demonstrated in a certain gray scale. There is a problem of insufficient data use in the classic LBP algorithm, so proposes a hierarchical LBP algorithm to solve the problem. In the new method, selects multiple threshold val-ues to serialize the images into LBP, then unions them and applies the improved HLBP algorithm to the face recognition. Based on YALE and ORL face databases, the experiments show that the recognition rate of HLBP algorithm is higher then that of classical LBP algorithm.
NETL
NSTL
维普
万方数据