Research On Screen Shooting Resilient Watermarking

With the development of electronic information technology,more and more files are stored and shared in the digital form,as a result,the copyright protection of digi-tal files becomes more and more important.As a branch of information hiding,digital watermarking can effectively realize copyright protection of digital files.By embed-ding identification messages in documents,images,audios and videos,the legitimate recipients can extract the embedded messages and effectively prove the copyright.In addition to copyright protection,watermarking can also be used for leak source trac-ing.However,traditional digital watermarking schemes cannot meet the requirements of leak source tracing nowadays,because the development of digital equipment greatly the way of information recording.The original information recording often needs to be done in hardware-contact way,by utilizing U-disk,hard-disk to copy files from the host.But the modern intelligent devices make information recording can simply re-alized only through mobile phone.By capturing the media files that displayed on the screen,efficient and high-quality information recording behavior can be achieved.The dissemination of such photos will cause a very serious information leakage problem.And the leaking source is hard to trace.Since the captured image will not affect the information signal but destroy the watermark signal which is normally designed for the electronic transmission channel,it is an urgent need to design a digital watermarking algorithm which can be robust to the screen-to-camera channel transmission.In order to meet such demand,this thesis carries out the research of screen-shooting resilient watermarking,so that the embedded watermark can be accurately extracted even after the screen-to-camera channel.Specifically,this thesis focuses on screen-shooting resilient watermarking algo-rithm in two levels:Firstly,based on the analysis of screen-to-camera channel distor-tion,a screen-shooting resilient image watermarking algorithm and a screen-shooting resilient text watermarking algorithm are designed.Then,in order to improve the two properties which digital watermarking algorithms focus on most(robustness and trans-parency),this thesis applies different technical means and designs corresponding algo-rithms from the embedding side and extraction side.The main works and contributions can be concluded as:1.Screen-shooting resilient watermarking for image carrierMost of the current image watermarking algorithms are focusing on the digital channel distortion,only a few works began to research on the cross media channel dis-tortion(print-scanning,print-shooting,screen-shooting).But the demand of leak source tracing of screen-shooting images has made screen-shooting robustness the most urgent requirement.To realize the screen-shooting robustness,this thesis analyze the special distortions caused by the screen-shooting process,including color distortion,lens distor-tion,light source distortion,and moire distortion.To resist the geometric deformation caused by lens distortion,this thesis proposed an intensity-based scale-invariant fea-ture transform(I-SIFT)algorithm which can accurately locate the embedding regions.As for the loss of image details caused by color distortion,light source distortion and moire distortion,this thesis put forward a DCT based small-size template algorithm to repeatedly embed the watermark into different regions,so that at least one complete information region can survive from distortions.At the extraction side,this thesis de-signed a cross-validation-based extraction algorithm to cope with repeated embedding.The validity and correctness of the extraction method are verified by hypothesis test-ing.Furthermore,to improve the robustness of keypoint,this thesis proposed a SIFT feature editing algorithm to enhance the intensity of the keypoints,based on which,the extraction accuracy and extraction speed can be greatly improved.The experimental re-sults show that compared with the previous schemes,the proposed algorithm provides significant improvement in robustness for screen-shooting process.2.Screen-shooting resilient watermarking for document carrierThis thesis designs a novel underpainting based screen-shooting resilient docu-ment watermarking algorithm for dealing with the leak source tracking problem.Com-pared with the traditional font based scheme,the proposed algorithm greatly release the restriction of language and screen-shooting robustness.Specifically,the designed underpainting contains three significant properties.1).Inconspicuousness.The water-marked underpainting is inconspicuous and it will not easily be maliciously attacked.2).Robustness.This thesis propose DCT-based watermark embedding algorithm and dis-tortion compensation based extracting algorithm,which make the watermark robust to camera shooting process.3).Autocorrelation.This thesis design the flip-based method to arrange the watermarked underpainting.So that a complete watermark region can be accurately located even if part of the document is recorded.Compared with previous watermarking algorithms,the proposed scheme guarantee the high visual quality as well as the robustness to the screen-shooting process.Besides,the screen-shooting experi-ments with multiple devices also show the good universality of the proposed algorithm.3.Robustness enhancement for screen-shooting resilient watermarkingTemplate based watermarking is a kind of watermarking scheme that can be used for cross media channel.In such scheme,the message is often represented by some dedicatedly designed templates,and then the message embedding process is carried out by additive operation with the templates and the host image.However,most of the traditional template based watermarking schemes use the handcrafted algorithm to extract the watermark features.Unless obvious templates are used,the robustness is not enough to resist the distortion,especially in screen-shooting task.At present,the success of deep learning in image processing tasks shows its strong feature extraction ability.Inspired by the powerful feature learning capacity of deep neural network,this thesis proposes a deep template-based watermarking algorithm,which can effectively enhance the extraction ability of template watermarking algorithm.Specifically,in the embedding side,based on the analysis of human visual system and robustness demands,this thesis proposes a series of criteria for template generation,and designs a template toexpress information based on such criteria.At the extraction side,this thesis proposes a two-stage neural network,which effectively guarantees the screen-shooting robust-ness.Various experiments show that the algorithm greatly improves the performance of extraction side,and the superiority of the screen-shooting robustness compared with traditional template based watermarking schemes.4.Transparency enhancement for screen-shooting resilient watermarkingRobustness and transparency are the most two important properties of digital wa-termarking schemes,but the existing template based watermarking algorithms cannot both satisfy the requirements of high transparency and high robustness.Therefore,this thesis proposes a scheme to enhance the transparency of the watermarking algorithm on the basis of ensuring the robustness.The enhancement of transparecny relys on the difference of observation frequency between human eyes and camera.Since human vision will behave "flicker fusion" phenomenon,that is,will observe the fused image with adjacent frame when facing high-frequency flicker images.But the camera,on the other hand will capture the single frame in exposure process.Therefore,using high frequency to display two images can effectively achieve the purpose that the human eye cannot observe but the camera can effectively record.Based on the analysis,this thesis proposes a digital watermarking scheme based on color decomposition principle and attention-guided neural networks.At the embedding side,a watermarking generation method based on Latin square scrambling is proposed to ensure the robustness.At the same time,to meeting the needs of transparency,a color decomposition based comple-mentary frame generation algorithm is performed which can effectively generate two complementary frames for further high-frequency displaying.To cope with the algo-rithm in the embedding side,this thesis proposes an attention-guided neural network structure in the extraction side,based on which,the non-distorted areas can be paid much attention.Thus,the accurate watermark extraction can be achieved.Extensive experiments show the transparency superiority of the proposed scheme and the strong robustness for screen shooting process.