基于DCT和线性回归的人脸识别

Face recognition based on DCT and linear regression

由于人脸图像常常因光照、姿态、表情变化及遮挡等因素的影响而具有非线性结构,在空间域直接使用线性方法就有其局限性。为此,提出了一种基于离散余弦变换和线性回归分类的人脸识别方法:通过离散余弦变换获取人脸图像的变换域特征,以减小光照、姿态变化等影响,然后再利用快速有效的线性回归方法得到识别结果。通过在几个常用人脸数据库上的测试结果表明,该方法在满足实时性的同时,能有效地增强面对这些问题的鲁棒性。

The performance of face recognition was limited when linear methods were used directly in spatial domain because face images own nonlinear structure when influenced by variations in illumination,pose,expression,and occlusion.To overcome the problem,this paper proposed a method based on discrete cosine transform and linear regression classification for face recognition.First it got the features in transform domain through discrete cosine transform,which purpose was to decrease the effect of variations in illumination,pose and so on,and employed then the fast and efficient linear regression method to produce the recognition results.The experiment results on several common face databases show that the proposed method is more robust in the case of real-time recognition is satisfied when confront these problems.