基于局部四值模式的人脸识别

Face recognition based on local quaternize pattern

局部二值模式(LBP)作为经典的纹理特征描述方法广泛应用于纹理分类和人脸识别等领域。然而现有相关算法仅利用周围一个圆形邻域的信息,没有充分利用周围邻域的信息。为此,提出一种利用不同圆形邻域之间的微分结构信息进行联合描述的特征描述子,从而能够更加充分地利用邻域信息。由于所提方法在圆形邻域上每个坐标处有4种不同可能的取值情况,因此将这种模型称为局部四值模式(LQP)。在通用的人脸识别数据库FERET上的大量实验证明了所提算法的有效性。

As a classic description method of texture features,local binary pattern（LBP）has been widely used in fields of texture classification and face recognition. However,the existing algorithms do not make full use of the surrounding spatial infor-mation but only exploit a circular neighborhood. To overcome the disadvantage,a novel descriptor which applies differential structure information between different circular neighborhoods to do joint description is proposed. It has four possible values at each coordinate in the circular neighborhood. Thus the model is called local quaternize pattern（LQP）. Extensive experiment re-sults on a popular face recognition dataset FERET show the effectiveness of the proposed method.