基于Trace变换的人脸特征提取技术研究*

为了提高人脸特征的稳定性和区分度,提出了一种基于Trace变换的人脸特征提取算法。算法通过几种不同的泛函函数对预处理后的人脸图像进行组合作用,得到该图像的一个Trace特征向量,从而建立了一种新的人脸特征表达方式。基于ORL人脸数据库的实验结果表明,该算法所提出的人脸特征对同一个人不同表情、不同光照条件下的图像变化能够保持较好的稳定性,同时对不同人的人脸图像具有较高的区分能力,在人脸识别的实际应用中是一种可行的方法。

Research of facial features extraction technology based on Trace transform

This paper proposed a novel facial feature extraction scheme and a new way for facial representation, which couldimprove the ability to discriminate different persons’ face images while also recognize same person’ s face images with differentexpressions or under different lighting conditions. It combined several functions together to process the preprocessed originalface images, which would produce feature vectors of the original face images. Experimental results on ORL face database provethat the facial feature extracted by this scheme can discriminate different persons’ face images while also recognize the sameperson’ s different face images. Therefore, this algorithm is feasible in practical applications of face recognition.