离散余弦变换和支持向量机相融合的人脸识别

FACE RECOGNITION BASED ON FUSING DISCRETE COSINE TRANSFORM AND SUPPORT VECTOR MACHINE

针对复杂条件下人脸识别性能低的难题,提出一种离散余弦变换和支持向量机相融合的人脸识别方法。首先将图像划分成子块,并采用对比度限制自适应直方图均衡算法对子块进行去噪处理;然后采用低频离散余弦变换系数来消除人脸图像中的光照变化;最后提取人脸特征,并采用支持向量机进行人脸识别。在多个人脸上进行仿真实验,结果表明,相比典型人脸识别方法,该方法不仅提高了人脸识别的正确率,同时减少了人脸识别时间,还提高了识别效率。

In light of the problem that in complex condition the face recognition has low performance,we propose a face recognition method which fuses the discrete cosine transform and SVM. First,it divides the face image into sub-blocks and uses the contrast limited adaptive histogram equalisation algorithm to conduct denoising process on sub-blocks; then it employs low frequency discrete cosine transform coefficients to eliminate illumination changes in face image,and finally extracts the face features,and uses SVM for face recognition.Simulation experiments are carried out on a couple of faces,results show that compared with other typical face recognition methods the proposed algorithm improves the face recognition accuracy,and also reduces the recognition time as well.