Firstly,in order to solve the problem that local binary pattern has small spatial support region and can only capture microstructure in face images.A novel face recognition approach which is multi-scale logarithmic difference face recognition based on local binary pattern(LBP)is proposed.Firstly,LBP operator is used to extract the texture feature of the face.Secondly,the LBP feature is used to extract the light invariant based on the Lambertian Reflectance Model.Then,the light invariant is used to obtain the multi-scale features according to the different distances,and the refined feature-map is obtained by the linear combination of multi-scale features.Finally,face recognition is performed using refined feature-map.Secondly,a new method based on sparse representation is proposed.Multiple scales logarithm is used to get the six scales feature-map,iteratively using AdaBoost algorithm to determine the weight of the six feature maps,and then linearly fused to obtain the refined feature-map,and finally use the sparse classification method to classify them. |