A new detector for JPEG decompressed bitmap identification.

In this paper, a new method for the identification of JPEG decompressed bitmap is presented. Consider the distribution of AC coefficients for the decompressed JPEG image, a key observation is that the local peaks of the distribution always appear in integer points whatever the quantization table is. We then propose to capture this effect by computing the frequency of AC coefficients falling in the set U-n is an element of z [n-1/4, n+1/4] which can cover all local peaks. One can prove that the frequency value is larger than 0.62. Meanwhile, for uncompressed image, empirical results show that the frequency value computed in the same way is just about 0.5. The validity of the proposed detector is then guaranteed. Experimental results show that the proposed method works well and is better than some prior arts.