Research On Blind JPEG Image Forensics

With the rapid development of Internet, the emergence of high resolution digital cameras, and powerful image editing software, doctored images can be found anywhere. It can greatly enrich the users live. People could see many beautiful images on magazine covers and commercial advertisement. But doctored images may also cause some problems. It will cause great social problems if they are used on news report or court. Therefore, the authenticity identification of digital image technology has been generally concerned. At present,the digital image that Image acquisition equipment outputed and the digital images on the Internet,most of them are stored using JPEG format. therefore the JPEG blind forensics has a very important practical value,However, the existing algorithm have more restrictive conditions,Away from practical use there is still a gap. In this context,this paper puts up two methods of detecting doctored images.Firstly, a method of blind JPEG image forensic via JPEG quantization coherence is proposed. Under the premise and suppose of different picture have different quality gene in the composite picture, this method detecting whether the picture were juggled by use of check the picture's quality gene consistent in the estimate processs. The algorithm computes quickly and efficiently, and its time domain diffcuity mainly centralize on estimateing Q value. The paper design experiment to validate its efficienty also, and the result shows that the algorithm can detect the picture efficiently whether it is original. The experiment also shows that if the origine pictue's quality gene and juggled picture's quality gene value discrepancy much than 5, the right ratio can pass 95%, howerve, less than 5 the right ratio can be affected.Sencondly, a method was put up based on JPEG picture's block domino offect. The method was put up based on the character when picture was composited, its 8×8 child block's variance, and the wave crest was attenuation and splited in the frequentcy domain. Accored to the block's variance in composited picture, the faked domain. was ascertained. This method eliminated the picture structure and vein noise by computeing g's each column power spectrum, and put them together. The pictures were compressed had periodic peak in its power spectrum, and the others had nothing. Analyse the local frequentcy domain power spectrum in higher frequentcy's flatness, use 32×64 glide block check the level block effects, and use 64×32 glide block check the vertical block effects. The results were multiplied to zoom out the discrepancy between the juggled domain and orginal domain. The result shows that the method is effectly to detect picture whether ture or not.Thirdly, compared to the block-based algorithm proposed in paper[65],our first algorithm both in accuracy and in terms of interference are superior to the paper[65] the proposed algorithm;our second algorithm is in terms of anti-interference in line with the algorithm proposed in paper[65]. However, in the determination of the distortion of the region, performance is slightly inferior to the paper[65] proposed algorithms.