A Research On Face Recognition Algorithm Under Complex Illumination

In recent years,with the rapid development of the field of biometrics,the related applications of face recognition in life have sprung up.However,in actual scenes,face recognition is always affected by various factors,including occlusion,expression,illumination,and posture.Studies have shown that the influence of illumination is the greatest among all factors and the difference of one person's images under different illumination is even larger than that of two different people.Complex illumination includes highlight,dark light,and shadows.Face recognition under complex illumination has great practical significance.This paper will study the face recognition under complex illumination to improve the image quality and improve the face recognition rate.It will mainly solve the problem of normalization of face image illumination and face shadow removal.The main research work is as follows:1.In view of the fact that most existing illumination preprocessing algorithms do not have the ability to process complex illumination images,an adaptive illumination preprocessing algorithm that can handle complex illumination is proposed.The algorithm can not only weaken the illumination effect to some extent to improve the image quality,but also effectively maintain the original potential features among the image pixels,which is conducive to subsequent feature extraction.2.Aiming at the problem that the shadow on the face image is difficult to remove and this problem will affect the feature extraction,an effective face shadow removal algorithm is proposed named adaptive shadow removal algorithm.The algorithm can accurately extract the edge of the light shadow and effectively remove the light shadow,which can improve the face recognition rate.3.Aiming at the problem that the illumination normalization algorithm still sensitive to illumination,a new illumination normalization method is proposed named modified Weber-Face.The algorithm not only can effectively extract the illumination insensitive features,but also effectively suppress the boundary marks at the light mutation,and can greatly improve the face recognition accuracy.The above work,through the multi-group comparison experiments on the three data sets,can prove the correctness and effectiveness of proposed algorithms.Thesealgorithms can be used for face recognition under complex illumination.The problem of interference of illumination on the accuracy of face recognition is solved perfectly.