Research And Application Of Digital Recaptured Image Algorithm

With the increasing convenience of digital image acquisition,its use as electronic evidence is also increasing.At the same time,a variety of powerful and easy-to-use image editing softwares are constantly appearing,making it easy to tamper with the image content."Seeing" may not be "real" and people are often plagued by "fake" photos.In many cases,tampering with the image may even cause political,diplomatic,commercial,legal,and other problems,which will cause a certain degree of negative impact on society.Therefore,the proof of image authenticity becomes more and more important.Digital image forensics technology has emerged with the utmost attention and has been highly valued in the academic and application fields.On the other hand,biometrics identification technology is gradually replacing traditional identity authentication methods.The unique features of each individual,such as fingerprints,facial images,and irises,can be used as the basis for identity authentication.And automatic identification based on fast,safe,and reliable human biological characteristics can bring huge economic and social benefits.However,with the gradual development and popularization of biometric identification systems,lawbreakers forge certain biometrics in some ways and pose as specific characters to achieve their own ulterior motives.Therefore,the judgment of the authenticity and primordial character of biological features has gradually been paid more and more attention by researchers.Digital image forensics as one of the branches of multimedia information security is a combination of image processing and information security.It refers to judge the originality,authenticity and completeness of the digital image content according to the image characteristics.In this paper,we study how to identify the two most commonly recaptured images: recaptured images taken from LCD screen and the recaptured images taken from printed pictures.The main research work of this paper includes:1)Build recaptured image databases.At present,there is no public standard recaptured image database,and the construction of recaptured image databases have important significance for the research of this work.We set up two image databases by ourselves: LCD screen recaptured image databases and Paper recaptured image databases.Among them,the LCD screen recaptured image databases includes 3000 original images and 3000 LCD screen recaptured images,and the Paper recaptured image databases includes 3000 original images and 3000 paper recaptured images.The image content and equipment in the two image databases are all diverse,and can meet the experimental requirements of this article in quantity.2)For the LCD screen recaptured images,according to the LCD screen recapture will cause the image edge blur and image color distortion,combined with the edge features of the gray image and the color characteristics of the color image,we proposed a LCD screen recaptured image detection algorithm.Using the ensemble classifier,the accuracy rate obtained in the LCD screen recaptured image database reaches 99.42%,and the accuracy rate obtained in the cross-image database reaches 87.05%.3)For the paper recaptured images,according to the paper texture remapping before and after the paper is recaptured,a paper recaptured image detection algorithm is proposed in this paper.The image features are constructed by extracting the texture features of gray images and the texture changes between color channels to identify whether the image is a paper recaptured image.Using the ensemble classifier,the accuracy rate obtained in the paper recaptured image database reaches 99.50%,and the accuracy rate obtained in the cross-image database reaches 87.85%.4)Combining the LCD screen recaptured detection algorithm into an actual face recognition system,a prototype for the identification of forged face recognition systems was implemented.This system mainly includes the registration and recognition of the human face.When the human face is identified,first identify whether the face is true or a fake face displayed on the electronic screen.If it is a real face,a registration ID number is given,and if the face is falsified,an error message is given.