The Design Of Image Pirate Detecting System Based On PRUN

The rapid development of picture and video editing technology enables people to edit pictures quickly and easily.This technology has two sides.On the one hand,the author team or the licensor can use the technology to improve the aesthetics and image quality of the work,and on the other hand,it prompts some people to tamper with unauthorized pictures for commercial purposes.It harms the original creator ’s interests and sometimes even changes the meaning of the picture itself.It is not conducive to creating an atmosphere that attaches importance to copyright,innovation and originality in China.Aiming at this problem,this thesis designs a method that can locate the tampered area of the picture and then detect the camera source of the tampered area and the background area respectively to achieve the purpose of copyright identification.Therefore,this thesis first investigates and summarizes the research background,significance,and status of domestic and foreign research on the two types of technologies: picture tamper detection and picture source detection.Secondly,some theories used in this thesis are briefly explained.Finally,some improvements are proposed.1.PRNU noise is relatively weak,and it is easily affected by image content and other sources of noise.Therefore,In the part of image noise reduction,we chose the spatial adaptive threshold filtering method instead of the winner filter,which can better eliminate the effect of the original content of the image on the PRNU noise.Because PRNU noise occupies a major part of the residual noise,we choose the main component of the residual noise as the PRNU noise.This can try to eliminate the impact of noise from other sources on the PRNU noise.2.Considering the structural characteristics of CFA,although half of the green channel in the CFA array affects the main component of PRNU noise,the PRNU noise of a single color channel does not better reflect the unique characteristics of each image sensor,so we Need to consider and make full use of the information of the three channels in the color image.Therefore,we extracted the PRNU noise matrix for all three color channels of the image and synthesized a PRNU noise matrix according to the CFA arrangement rules.3.Using the DCT coefficient histogram directly in the step of calculating the posterior probability map in the image tampering detection may result in low accuracy.Therefore,we first need to estimate the quality factor of the first quantized compression of the image,and then calculate the posterior probability map based on the double quantization periodic function and the DCT coefficient histogram to improve the accuracy of tamper detection.