Research And System Implementation Of Eyeglasses Removal Method For Frontal Face Image

With the urgent needs for social security issues,security technology has become a hot topic of social concern.Among them,face recognition is one of the important researches on security technology,mainly on the research of biometric technology.In the process of recognition,there is a problem that the face recognition accuracy is low and the robustness is poor,mainly because the face of the person is disturbed by factors such as illumination,facial gestures,facial expressions,and facial occlusion.However,it is easy to find that dark-rimmed glasses are the most common obstruction in facial obstruction,which seriously damages the face recognition accuracy.Therefore,this paper studies the method of removing glasses in frontal face images.The main research contents include:A facial glasses removal method based on edge information projection and improved TV repair model is proposed.Firstly,this paper determines whether there are glasses in the face according to the edge information of the nasal bridge region;secondly,locates the area where the glasses are based on the maximum position coordinates projected by the edge information;then,in the occluded area of the locating glasses,the glasses template is extracted by automatically acquiring the region growing method of the seed area points,and in order to avoid extracting the non-glasses area,the eye area is covered by the ellipse;Finally,,according to the extracted glasses template image,the diffusion repair of glasses occlusion area is performed by the improved TV repair model,and the removal of the glasses in the face image is completed.The experimental results show that the proposed method based on edge information projection and improved TV repair model can effectively remove the glasses in the face.Compared with the removal effect of the TV repair model,the visual effect of the method is better.An automatic glasses removal method based on feature reconstruction of convolutional neural networks is proposed.Firstly,this paper searches for wearing glasses face image and no glasses face image with similar face attributes as the image set uses the K-nearest neighbor algorithm;Secondly,the image sets are respectively mapped to the feature space of the network to calculate their average feature values;and then calculate the difference between the two to obtain the glasses attribute;Finally,the obtained glasses attributes and the glasses face image mapped to the feature space are subjected to subtraction operation,the attribute removal of the glasses is completed,the face features need to reverse mapping back to the pixel space,this paper completes the reconstruction of the pixel image according to the design of the objective function.Based on the above work,this paper designed a frontal face glasses removal system with Matlab.The system mainly includes the glasses removal method based on edge information projection and improved TV repair model and the glasses removal effect and performance evaluation of other repair models.