| Image compression is one of the most widely used techniques among digital image processing field, and JPEG and JPEG2000 are having been utilized in our life for a long time. However, compression will cover the digital tampering operation meanwhile. Hence under some circumstance, such like news publishing or proposing court evidences, images are requested to be original without any process, even for widely used image compressing operation. But it is rather tedious and difficulty for human to do this job, so automatic detection of image compression is important.Compressed images will show special correlation characters, and it can be as the evidence for compression detecting. However compression presents similar correlation features with interpolation, thus the difficulty of compression detection is to differentiate it with interpolation detection. Powerful computers, high-resolution cameras, and sophisticated photo-processing software packages are making digital forgery convenient for average people. Our trust on image authenticity may become a thing of the past. Passive authentication has been proposed to solve these problems recently. Compared with watermark it needn't put anything into the original images. Alin C. Popescu and Hany Farid gave two algorithms to exposing images forgeries by detecting traces of re-sampling and CFA interpolation, but its robustness to JPEG and JPEG2000 is underperformed.In this paper we thoroughly analyze correlation features of JPEG and JPEG2000 compressed images, then we present an algorithm for the automatic detection of JPEG and JPEG2000 compression based on EM algorithm. Preliminary experimental results show that our method is effective and accurate with some limitation. |
PDF Full Text Request