传统协同表达分类( CRC)算法因直接使用原始样本构造非传统字典,容易受到样本维度、光照和姿态变化等因素的影响.该文在协同表达框架基础上,提出了一种新的利用分块加权局部二值特征( LBP)直方图向量构造解析字典的协同表达人脸分类方法.首先通过分块加权方法优化LBP算子提取的纹理特征,然后采用解析字典学习方法将样本数据投影到稀疏系数空间,并使用协同表达方法重构测试样本,完成样本分类.与已有算法相比,该文算法的实验结果较好. ORL和LFW数据库上的实验结果证明了该文方法的有效性. cation method based on CRC is proposed by using a set of the weighted block-based local binary pattern( LBP ) histogram vectors to construct an analytic dictionary. A block weighted method is presented to optimize the texture features extracted from the LBP. Secondly, the samples are projected into a sparse coefficient space,which is constructed by an analytic dictionary model. The final goal is to perform the robust face classification using the proposed hybrid method. Experimental results conducted on the ORL and the LFW face databases demonstrate that the proposed method has the desirable classification performance.