利用人眼分形码距离进行快速人脸识别

在利用人脸分形码距离进行识别时,需要大量的时间对人脸库中每张人脸图像进行迭代与距离运算.为克服这一缺点,本文提出了用水平方向高频子带来定位眼睛并将其从人脸中抽取出来,进一步提出了基于人眼分形码距离的人脸快速识别算法.利用该算法,可去掉大部分人眼分形码距离较大的图像,从识别时间复杂性分析,本文算法所需时间主要与人眼大小以及用于最后识别的图像数目有关.在ORL和YALE两个人脸库上的实验结果表明,与本征脸方法和直接利用人脸分形码距离方法比较,在用于最后识别的图像数目占人脸库中人脸总数的20%左右时,本文算法可使平均识别率保持在约90%,与其它方法基本持平.

Fast face recognition using the distance of fractal codes for the eyes

Since recognizing the distance of fractal codes of face needs a large mount of time for iterative operation and calculation of the distance of each face in the face database, this paper presents the method for locating the eyes with high frequency subband in horizontal direction and extracting them from the face. Moreover, a fast face recognition algorithm based on the distance of fractal codes for eyes is proposed. Most images with large distance of fractal codes for eyes can be excluded by using this algorithm. The time needed by the proposed algorithm mainly depends on the size of eyes and the number of the final recognized images. The experimental results in ORL and YALE face databases show that, the average recognition rate retains about 90% while the number of the final recognized images is about 20% of the total number of the face images in the face database. Compared with the eigenface method and the method directly using the distance of fractal codes for face, the recognition rate is the same but the spent time is much less.