Based On Face Fusion And Singular Value Decomposition Face Image Privacy Protection

With the rapid development of digital technology,the acquisition,distribution and sharing of digital images are becoming more and more popular in information transmission and communication.At the same time,automatic digital image processing and analysis technology has been widely developed,which provides convenience for people to identify,analyze and understand images.However,it also facilitates potential attacks by malicious users,which may bring security risks.Especially for face images,malicious users can obtain various biometric information from the Shared face images.Therefore,facial privacy protection is particularly important.An ideal privacy protection method for face images should be able to prevent malicious automatic analysis and minimize the effect on the effectiveness of images.Face privacy protection in this paper is mainly based on image editing method.The existing privacy protection methods based on image editing can be divided into full-face privacy protection method and face roi-based privacy protection method.For the former,it edits the entire face image area to achieve privacy protection.Face fusion is one of the typical methods,which can be used to confuse face features in face images.However,with the continuous development of face recognition technology,the privacy protection capability of this method is limited.In the latter,the region selection stage is added.For example,face component fusion method can select face eyes,nose and mouth as the modification and adjustment part.It retains the non-component regions of the face and has a high privacy protection capability.However,when all the components are replaced,the visual quality of the face image decreases significantly.In view of the above problems,this paper does the following research work:First,a face image privacy protection method based on face fusion and singular value decomposition is proposed.Firstly,a face image which is similar to the input source image is found in the face image database and preprocessed.The two images were then fused with faces;Then,singular value decomposition is performed on the fused face image.This step can modify or remove the principal components of the face image in the source image and further modify the biometric information of the face image in the source image.Finally,in order to balance the privacy protection and subjective recognition of face images,the image obtained by singular value decomposition can be weighted with the image after face fusion.Experimental results show that the proposed scheme is effective in privacy protection and subjective recognition of face images.Second,a face image privacy protection method based on face component fusion and singular value decomposition is proposed.The first scheme ignores the importance of face components to face images.Therefore,on the basis of the first work,we no longer process the whole face image.The second scheme retains the contour of the face image of the source image,and more specifically processes the face components that contain most of the face biometric features.The face components selected in this scheme are eyes,nose and mouth.Singular value decomposition is carried out on these face components respectively,which can better balance the privacy protection ability of face images and subjective recognition of face images.The experiment shows that this scheme can not only retain some information of the face source image,but also avoid the distortion of the image,and achieve better privacy protection.