Reversible Face Privacy Protection Based On Image Geometry Deformation

In recent years,with the rapid spread and development of the Internet and mobile devices,images have become an important means for people to record and share their lives.More and more image data appear and are stored on the Internet.With the advent of the era of brushing,face verification has become an important means of daily identity verification.At the same time,more and more people are paying attention to the security of photos containing faces uploaded to the Internet.Therefore,it is urgent to take measures to protect the privacy of faces in images and prevent the leakage and abuse of face privacy.Face privacy protection is mainly to remove the prominent identity features contained in the face of the image,and to eliminate the connection between the face and its owner.This thesis focuses on the face privacy protection algorithm,summarizes the fuzzy processing and encryption processing algorithms commonly used in face privacy protection,and analyzes the advantages and disadvantages of the existing algorithms.Aiming at the problems of image visual effects,image usability and the irreversible recovery in the face privacy protection algorithm based on fuzzy processing and encryption processing,this thesis designs a reversible face privacy protection algorithm based on global image geometric deformation.The algorithm firstly performs block mapping on the global face region,obtains the target deformation shape of each block,and then performs reversible geometric deformation on each small block,records the missing pixel information and the position of the interpolated pixel region during deformation.In the process of deformation,the upper four bits of the missing information are filled in the lower four bits of the interpolation region to obtain the image after the face privacy protection.Finally,the lower four bits information and the face coordinates information are hidden in the entire image after the face privacy protection by the reversible information hiding technique.The experimental results show that the algorithm can reversible recover the original image and maintain the security of face privacy protection while maintaining image visual effects and usability.In order to improve the visual effects of face after face privacy protection,and improve the robustness of the algorithm in lossless recovery,this thesis designs a reversible face privacy protection algorithm based on image geometric deformation of key regions.The algorithm firstly detects the key points in the face,and then predicts the key regions of the face through key coordinates.The reversible geometric deformation operation is performed on the key regions of the face to complete the privacy protection of the face,which effectively maintains the contour of the face.At the same time,the test method of face identity similarity is improved,and a cross-face contrast analysis structure is designed,which can more effectively evaluation the security performance after face privacy protection.Experimental results show that compared to the reversible face privacy protection algorithm based on global image geometric deformation,the face detection confidence is improved,the facial identity feature similarity is reduced,the reversible recovery of the image is guaranteed,and the quality of the restored image is improved after the algorithm is used for face privacy protection.The algorithm achieves better image visual effects based on ensuring image reversible recovery,and the image availability and privacy protection security are greatly guaranteed.Finally,in order to visualize the flow and results of the algorithm,a demonstration system was designed to visualize the presentation.