中文摘要：

针对现有人脸识别系统对模糊人脸图像的识别无法达到理想的效果,提出一种基于局部相位量化（Local Phase Quantization,LPQ）和Fisherfaces进行模糊人脸识别方法。首先采用LPQ算子提取分块模糊人脸灰度图像的LPQ直方图序列（LPQHS）,然后对采样后的特征运用Fisherfaces方法进行特征子空间选择,最后通过最近邻分类准则进行人脸识别。该算法增强了提取模糊人脸纹理信息的有效性,使训练数据量大幅度降低,并且图像特征向量的维数与原始图像大小无关。在Yale和AR形成的模糊人脸数据库上的实验表明,该方法具有较高的识别率。

英文摘要：

Aiming at what the existing face recognition systems identifying blurred face image cannot achieve desired results,a blurred face recognition method is proposed based on local phase quantization（ LPQ） and Fisherfaces. Firstly,the LPQ histogram sequence（ LPQHS） is extracted from block grey-level face images by the LPQ operator. Then the Fisherfaces based feature selection method is applied to extract feature subspace.Finally,the recognition is performed using a nearest neighbor principle. The algorithm enhances the effectiveness of extracting the blurred face texture information,so that the amount of training data is greatly reduced,and the dimension of feature vector of the image and the original image are independent of the size of a blurred face. The simulation experiments illustrate that the proposed method obtains a better recognition rate on both AR and YALE face database comparing to classical methods.