Abstract: In order to improve the performance of face recognition under non-uniform illumination conditions, a face recognition method based on Patterns of Monogenic Oriented Magnitudes (PMOM) was proposed. Firstly, multi-scale monogenic filter was used to get monogenic magnitude maps and orientation maps of a face image. Secondly, a new operator named PMOM was proposed to decompose the monogenic orientation and magnitude into several PMOM maps by accumulating local energy along several orientations, then Local Binary Pattern (LBP) was used to get LBP feature map from each PMOM map. Finally, LBP feature maps were divided into several blocks, and the concatenated histogram calculated over each block was used as the face feature. The experimental results on the CAS-PEAL and the YALE-B face databases show that the proposed approach improves the performance significantly for the image face with illumination variations. Other advantages of our approach include its simplicity and generality. Its parameter setting is simple and does not require any training steps or lighting assumption and can be implemented easily.

Key words: face recognition, monogenic filter, lighting, magnitude, orientation

摘要： 为了提高在复杂光照下的人脸识别率,提出了一种基于单演定向幅值模式的人脸识别算法。首先,用多尺度的单演滤波器提取图像的单演幅度和方向信息;然后,用一种新的单演定向幅值模式(PMOM)算子将同一尺度下的幅度和相位信息分解为多张定向幅值模式图,再用局部二值模式(LBP)算子提取每一个PMOM模式图的LBP特征图;最后,将每张LBP特征图分块,计算每一块的直方图,并将所有块的直方图串联后作为最终的人脸表示。在CAS-PEAL人脸库和YALE-B人脸库上的实验结果表明,该算法可以显著提高光照变化人脸图像的识率。另外,该算法参数设置简单,而且无需任何训练过程也无需对光照条件进行估计,因而具有简单、通用性好的优点

关键词: 人脸识别, 单演滤波, 光照, 幅值, 方向