基于Gabor-PCA的人脸遮挡区域重建算法  

GABOR-PCA BASED RECONSTRUCTION ALGORITHM OF OCCLUDED FACE REGION

作　　者：黄孝平[1] 

机构地区：[1]南宁学院,广西南宁530200

出　　处：《计算机应用与软件》2016年第3期210-213,283,共5页Computer Applications and Software

基　　金：2012广西高校特色专业及课程一体化建设项目(GXTSZY305)

摘　　要：针对遮挡人脸重建和识别问题,提出一种基于Gabor滤波和主成分分析相融合的人脸遮挡区重建算法。首先通过构建5维8方向的Gabor直方图信息分类器,从人脸库中选择Gabor直方图信息与待重建原始人脸图像在外形轮廓等粗信息更为接近的图像组成训练样本集,然后采用主成分分析对奇异值分解重建协方差矩阵进行投影形成重建图像,最后进行仿真实验。结果表明,该算法可以得到平滑自然的重建人脸图像,而且具有较强的适应性。Aiming at the occluded face reconstruction and recognition problem,we proposed a Gabor and PCA fusion-based reconstruction algorithm for occluded region of face. First,by constructing Gabor histogram information classifier with five dimensionalities and eight directions we selected the images from database which have the Gabor histogram information much close to the coarse information such as the outline of the primitive face images to be reconstructed to form training sample set,and then adopted PCA to decompose singular value and to reconstruct covariance matrix as well as to project to from the reconstructed image,and finally the simulation experiment was carried out.Results showed that the proposed algorithm could obtain smooth and natural reconstructed face images,and had stronger adaptability.

关 键 词：人脸重建 训练样本 主成分分析 遮挡人脸 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]