【摘要】 针对基于局部方向模式LDP(Local Directional Pattern)人脸识别方法识别率较低的问题,提出一种基于改进LDP的人脸识别方法。该方法对LDP编码进行方向归一化处理,选取结构对比信息对各LDP人脸分块进行加权处理,提取人脸的LDP直方图特征,最后计算加权卡方距离得到人脸识别的结果。实验结果表明,该方法应用于FERET和CMU PIE人脸图像库,相对于PCA、LBP和LDP算法,识别率均有了明显的提高。更多还原

【Abstract】 In light of the problem that the recognition rate of face recognition method based on local directional pattern( LDP) is unsatisfactory,in this paper we propose a method for face recognition which is based on improved LDP. The method normalises the directions of LDP codes,selects the structure contrast information to carry out the weighted processing on each face image sub-block of LDP,and extracts the LDP histogram features of face image,at last calculates the weighted chi-square distance to get the face recognition results. Experimental results show that applying the proposed method to FERET and CMU PIE datasets,the face recognition rate improves significantly compared with PCA,LBP and the original LDP algorithm.更多还原