基于Sobel与LDP特征的人脸识别算法研究

Research on Face Recognition Algorithm Based on Sobel and LDP Features

针对人脸识别过程中采集到的人脸图像环境不同、存在非一致性光照和随机噪声等问题,提出一种基于的Sobel算子与LDP特征值的改进人脸识别算法。首先对采集到的人脸图像,依次利用加权平均值算法、改进的Sobel算子进行灰度化和图像增强,再通过改进的局部方向模式得到人脸特征直方图,最后与人脸库中的人脸特征直方图进行匹配。在yale人脸库上进行了实验,实验结果表明:这种方法确实可以结合两种改进算法的优势,同时具备较短识别时间和良好的识别率。

Aiming at the problems of different face image environment collected during face recognition,inconsistent illumination and random noise,an improved face recognition algorithm based on Sobel operator and LDP eigenvalues is proposed. First,for the collected face images,the weighted average algorithm and the improved Sobel operator are used to perform grayscale and image enhancement,and then the face feature histogram is obtained through the improved local direction mode. Finally,it is matched with the face feature histogram in face database. Experiments are conducted on the Yale face database. The experimental results show that this method can indeed combine the advantages of the two improved algorithms,while having a shorter recognition time and a good recognition rate.