Research On Camera Micro-motion Parameter Estimation Based On Video PRNU Noise

With the rapid development and popularization of smart phones and short video platforms,everyone has a mobile phone becomes a matter of course.People can already shoot videos anytime and anywhere,and even upload videos to the Internet anytime,anywhere.The rapid development of short video platforms has accelerated the transmission of information and has become one of the most popular methods of information transmission at this stage.However,the convenience of shooting,uploading and publishing personal videos will also bring disadvantages.For example,videos containing illegal content will increase exponentially.In addition,publishers of illegal videos usually use the network’s privacy protection mechanism for users to commit crimes anonymously,which brings great difficulties for forensic doctors to obtain evidence.Therefore,the source identification of multimedia data(image and video)has always been one of the most concerned problems for researchers in the field of multimedia forensics,especially the research on the trace of camera micro motion in the process of video acquisition.This paper mainly studies the current status and development of source identification technology for anonymous video,and discusses the problems and solutions encountered in the research process of source identification technology at this stage.At the same time,the feasibility of tracing back to the micro motion of the camera is studied and analyzed.First of all,this article expounds the development background,social significance of multimedia forensics technology and the current research status at home and abroad.Secondly,this article introduces the relatively new and effective algorithms at this stage,and summarizes the relevant theoretical knowledge involved.Finally,aiming at the problem of source identification of anonymous videos,an improved camera micro motion matching algorithm based on Photo-Response Non-Uniformity(PRNU)noise is proposed.According to the needs of the subsequent research of the algorithm,a new data set on images and videos was created.Then,on the basis of the new data set,the source identification algorithm which can only be traced back to the video source device is extended to the stage where the micro motion parameters of the camera can be estimated.The innovations of this article are mainly embodied in three parts: the first is that this article proposes a new image and video data set,which contains images and videos taken by relatively new equipment at this stage,and innovatively records the personal information of the photographer to facilitate subsequent research on the relationship between the photographer and the video.The second is the camera micro motion matching algorithm based on video PRNU noise,which discusses the adverse effect of the video stabilization function of the new shooting equipment on the video source identification algorithm and its three-dimensional characteristics,and then it is concluded that the PRNU noise of the test video needs to undergo a three-dimensional geometric transformation to promote more accurate matching.The performance evaluation of this method by using the standard database shows that the identification result of this method is better than the existing methods in accuracy.Third,on the basis of the above-mentioned improved algorithm and new data set,this paper proposes a camera micro-motion parameter estimation algorithm based on video PRNU noise.It is known that the goal of equipment stabilization technology is to reduce the impact of the photographer’s non-subjective conscious control of hand shaking on video quality.When the method can more accurately simulate the motion parameters of the hand shaking behavior,the motion parameters of camera experience can be estimated during shooting,and can further analyze the motion state of the photographer,which is of great significance to the analysis and reasoning of the case.