Logarithmic Total Variation Correction Method For Face Recognition

To meet the demands of the new age,biometrics technology has become one of the hot topics in the field of pattern recognition in recent decades.In the field of biometrics,the most convenient and widely used identification method is face recognition,but due to technical limitations,the proportion of face recognition in the market has been ranked second,only lower than fingerprint recognition.Eigenface method is one of the main methods of face recognition feature extrac-tion.The basic idea is to process the original face image,remove irrelevant redundant information,obtain eigenface image which is mainly contain the face features to par-ticipate in subsequent matching and identification afterward.The extraction effect is mainly affected by light condition,image background,posture,expression,and oc-clusion.In the existing eigenface method,such as self-quotient image method(SQI),logarithmic total variation model method(LTV),etc.These methods have good ef-fects on eliminating illumination,but are vulnerable to changes in posture,expression,and occlusion,resulting in low recognition rate.For face databases that include mul-tiple effects of illumination,image background,gestures,expressions,and occlusion,the researchers proposed collaborative representation classification(CRC),sparse rep-resentation classification(SRC),wavelet transform(DWT),and a joint of pixel-level and feature-level wavelet transform fusion method(TWSBF)is used to extract image features,but all of these methods still have space to improve.This article mainly discusses the eigenface extraction method of face image.The research idea is to exclude the interference of background and other information in the image,so that the face outline and facial features in the given face image are highlight-ed to participate in the subsequent recognition and matching.Because of the effect of feature face extraction based on logarithmic total variational model(LTV)is not obvious to facial features,the recognition rate under the influence of posture,expres-sion,and occlusion factors is quite low,This paper proposes a logarithm total variation correction model(CLTV)for face feature extraction,which guarantees the existence and optimality of the model solution.The model is solved by the alternating direc-tion method.The obtained feature image has better performance than other eigenface methods.The facial recognition simulation experiment is performed on several public face databases.The results show that the method has a better recognition effect than other eigenface methods and some non-feature face methods.Based on this,this paper considers the influence of the nuances of large-area background pixels on the recog-nition and matching,proposes a binarization processing method for eigenface images,which further improves the recognition rate.