基于位平面图像的人脸特征抽取

Face feature extraction based on bit-planes image

提出一种新的基于位平面图像的特征抽取方法.该方法通过对原始图像进行"位切片",将1幅图像分解为8幅位平面图像,然后针对不同的位平面图像的特点,对不同的单幅位平面图像和合成的位平面图像进行特征抽取,为从复杂的人脸图像中抽取出有效的鉴别特征提供了一种有效措施.ORL标准人脸库中的实验数据验证了该方法的有效性.

In this paper, a novel method for feature extraction based on ＂bit-planes image＂ is developed. It slices the image at eight different planes （bit-planes）. By means of the difference of bit-planes images, feature extraction is done for different single and combined bit-planes images. The proposed method affords a new method for extracting effective feature from complexity image such as faces. Experimental results on ORL face database verify the effectiveness of the proposed method.