光照鲁棒的人脸识别

传统的人脸识别方法对待识别人脸图像的质量要求较高，而且要求所采集的人脸图像的光照情况与人脸训练库的光照情况的差异不能太大，这就限制了人脸识别系统运行的环境条件，从而限制了人脸识别的应用。为了降低人脸识别对环境条件的要求，真服光照对人脸识别的影响。本文分析了人脸图像的幅频特性和相频特性，提出了频域光照归一化的人脸识别方法，使得对任何光照条件下采集的图像经过归一化后，光照情况与训练库中的图像完全相同，同时保留了人脸的可区分性。因为人脸之间差异的信息量一般较少，故本文运用最小非零特征向量作为人脸特征，通过实验仿真，与传统方法相比本文人脸识别方法对光照变化具有鲁棒性。

Illumination robust face recognition

The traditional face recognition method has high requirement to the face image to be recognized and require that there are httle illumination differences between the face image acquired and the image in the training database, which restrict the environmental condition in which the face recognition system is operated, thus restrict the application of face recogntion. In order to lessen the requirement of environmental condition in the face recognition and overcome the effect of illumination to the face recognition, the paper analyzes the amplitude - frequency characteristic and phase - frequency characteristic and puts forward the illumination normalized face recognition method in the frequency domain. By normalization, the illumination between the image acquired and the image in the training database is identical and also the distinguishable property of face image is preserved. Generally,the information of the differences among the face images is less, so this paper considers the minimum non - zero eigenvector as the face feature. By the experiment simulation, compared to the traditional face recognition method the method put forward in this paper is robust to illumination change.