Face Recognition Based On Image Texture Feature Extraction

Face recognition is a research hotspot in the field of pattern recognition,which has been widely used in many fields.Face feature extraction is one of the key steps in the face recognition system.Facial texture features are highly discriminative,so extracting the texture features of face images often can get a good recognition effect.In the last decades of research,many effective methods of texture feature extraction have been created.Generally,these methods can be divided into 4categories: structural method,statistical method,model method,signal processing method,and so on.The extraction of the texture feature dimension,the ability of discrimination,robustness and computation are the criteria to measure the performance of the texture feature.Local binary pattern(LBP)is a kind of computational complexity.In machine vision,it has shown good performance,and is widely used in fingerprint recognition,face recognition,image retrieval and other fields.Local ternary pattern(LTP)is an extension of the LBP algorithm,by adding a threshold to a certain extent to resist the influence of noise.However,the LTP method and LBP method can only describe the texture characteristics effectively in a certain degree of thickness.In this paper,a LTP method based on multi scale pixel block weighting is proposed.According to its own situation,the method can describe the texture feature from coarse to fine,and the PCA algorithm is used to reduce the dimension of the improved LTP operator.On this basis,according to the LTP method,the threshold selection is often more difficult problem,proposed a method to automatically select the LTP threshold.Finally,because the LTP method is not effective in the global description of the image information,a fusion of DCT and improved LTP texture feature extraction method is proposed.Face image belongs to a kind of irregular texture image.The proposed method can better describe the human face.In the ORL face and YALE database,this algorithm was validated.The experimental results by using the method of facial feature extraction combined with nearest neighbor classification method has high recognition rate.