复杂光照条件下的人脸识别方法

Face recognition under complex illumination environment

在人脸识别增加真实性的研究中,为了提高在光照条件变化时人脸图像的识别率,并加快运行速度,提出了一种基于小波变换域的光照处理与识别方法。由于光照对低频信息的影响较小,且低频信息在人脸识别中起到最主要作用,通过对人脸的低频逼近图像进行光照处理,采用局部二元模式来表征光照处理后的低频图像,将得到的局部二元模式特征作为人脸的鉴别特征用于分类与识别。根据YaleB、Extended YaleB人脸库的实验结果表明,在复杂的光照条件下识别率高达96%,与传统方法相比,取得了更好的识别结果。

In order to increase the recognition rate under illumination variations,a processing of illumination and face recognition method based on wavelet transform domain is proposed.Because the illumination does not significantly affect the low-frequency information which plays an important role in face recognition,the processing of illumination is applied to the approximate images,and the Local Binary Pattern method is performed to characterize the approximate images after the processing of illuminition,and the Local Binary Pattern feature is used as the face description for classification and recognition.The experiment results on YaleB and Extended YaleB face database show that the proposed method can achieve high face recognition rate up to 96% under complex illuminations,and it is much better than the traditional method.