一种改进的特征脸方法

RECOMMENDATION OF AN IMPROVED EIGENFACE METHOD

在实际应用中,我们发现,已有的一些人脸识别方法对于每人一个样本的识别来说,效果不太理想。鉴于此,本文将在传统的特征脸方法理论基础上提出一种改进的特征脸方法——特征半脸方法。所谓特征半脸方法,就是把人脸图像分成上下两个部分,分别应用特征脸方法,最后在识别计算距离时上部采用较大的权重,下部采用较小的权重,求得综合距离最小的人脸图像序号,从而完成人脸识别的方法。我们把特征脸和特征半脸方法进行了对比实验,结果表明:新的特征半脸方法优于传统的特征脸方法。

In practice, we find that by the application of a human face database with single sample for each person, the results of old face recognition methods are not so good. After an introduction of the basal theory of eigenface method,we will introduce a novel method-half- eigenface, which is an improvement of eigenface method. Half- eigenface method is a method that through separating the face image into two parts: the upper and the lower,then they are processed respectively with eigenface method. Ultimately, when we calculate the synthetical distance by assigning the upper part to a bigger weight than the lower part,we take the face image with the smallest distance as the result. We have compared the novel half- eigenface method with the eigenface method through experiment, and the result shows that half- eigenface is better than eigenface.