Detecting Doctored Images Based On Double JPEG 2000 Compression

With the rapid and extensive development of computer and network technique, the emergence of high resolution digital caneras,and sophisticated photo-editing software (such as Adobe Photoshop,Microsoft Photo Editor,CorelDRAW), doctored images can be found anywhere,moreover,they are hard to distinguish from authentic photographs.These forgeries, if used in the mass media or courts of law, can have an important impact on our society. It is very important to detect the doctored images. Therefore the detection of doctored images has became a research focus in the information field.This dissertation summarizes the research status of tamper detection, and based on the good performance and development prospects of JPEG 2000 image compression standard, this dissertation proposes a method of detecting doctored double JPEG 2000 images. The dissertation analyses the statistics properties of double JPEG 2000 compression, and the artifacts in the Fourier transforms of subband DWT coefficients histogram, that is,compared with the first compression, stronger peaks are visible in the mid and high frequencies. Based on the characteristics of Wavalet transform which can be used to locate the singularity (peaks), then,according to the range(the difference of maximum and minimum),which is induced by the peaks,we propose a method of detecting double JPEG 2000 compression. The dissertation points out that it is divisible of doctored blocks and undoctored blocks, and so a simple and effective method is used to locate the doctored part.The experiments show that this algorithm not only can detect double JPEG 2000 compression, but also can accurately locate the doctored part.