A Human Face Classification Approach on Local Color Binary Patterns

Aiming at low classification rate of local texture features for illumination human face classification,this paper proposed a local color binary patterns method for texture feature extraction and presentation.We first compute the norm of color vector between multi-channels to reduce quantify errors caused by extracting LBP features from each channel individually.Then,we extract the pixel color angular between each two channels to reduce illumination influence.Given test human face image dataset,the face is classified by SVM.The experimental results on standard dataset Color FERET and XM2VTSDB demonstrate the effectiveness of our method.