Blind Forensic Of Color Forgeries Based On The JPEG Compression Characteristic

With the increasing popularity of imaging device such as digital camera, the digital image has become the principal ways to acquire and disseminate information. But rapid advancement in image processing and editing software has made it much easier to tamper images. The digital forgeries which flood on the internet and all kinds of media have affected people's daily lives. So the blind forensics on the purpose of identifying and detecting the tampered images becomes a research hotspot currently. Because the majority of images on the internet and in the cameras are JPEG format and forged images are always stored in JPEG format, the detecting of the JPEG compressed forgeries has important academic value and practical significance.Firstly, we summarized the principle and characteristic of current blind forensics and analyzed the status and deficiency of these means towards JPEG images. By researching the generation of nature images, the JPEG compression and the primary tampering means, a general JPEG tampered model is given as the subject investigated in the paper.By the analysis of double quantization characteristic in JPEG image in depth, the point that different double quantization processes could make different distortion in images is given. An approach, compressing the images again and exploring the distortion of different areas, is proposed. Because different regions have disparate quantization history, the tampered areas can be distinguished from the background according to various distortions in the recompressed image. All the detecting results of the three color components are synthesized to confirm the tampered areas when detecting the color forgeries. The simulation results show that the approach is effective in detecting the color forgeries and is better than the way based solely on a component.Local block artifact in JPEG image is defined and measure formula is given on the basis of analyzing the overall block artifact of JPEG images. According to the features of tampering model, the detecting approach that concludes local blocking artifact of different regions in detected images using different sliding blocks along the horizontal and vertical direction is given. The detection of doctored region is realized via checking significant differences of block artifact in different areas. Then the improved method is presented in order to search the optimum condition for different doctored images.