Tamper Detection Method for Clipped Double JPEG Compression Image
Gong Zhenzhen, Niu Shaozhang
Beijing Key Lab of Intelligent Telecommunication
Software and Multimedia
Beijing University of Posts and Telecommunications
Beijing, China
gongzhenzhen2013@sina.cn, szniu@bupt.edu.cn
Han Hongli
School of Science
Beijing University of Posts and Telecommunications
Beijing, China
hlhan@bupt.edu.cn
Abstract—with the development of digital image processing
technique the credibility of digital image has been questioned.
This paper proposes a new tamper detection method for
clipped double JPEG compression image based on the
statistical characteristics of DCT coefficients and block effect
caused by double JPEG compression. The offset of the two
DCT grids can be obtained by computing the loss of
information. Then we will trim the image according to the
offset, and it will be available to estimate the primary
compression quality factor. The tampered area can be located
according to the difference of DCT coefficients. Experiments
results show that the proposed method is robust and effective
in detecting the clipped double JPEG compression image. And
the method is also applicable to the unclipped double JPEG
compression image.
Keywords-image clipping; double JPEG compression; loss of
information; recompression offset
I. INTRODUCTION
The popularization of digital cameras and digital image
processing software make it an easy job to do image
tampering, and the tampering methods are various. As a
result, the authenticity of the digital pictures is often
questioned by public. It has impacts on the law enforcement,
the media, national security and so on. Digital images are
easy to be tampered while difficult to be detected, so the
study of image forensics detection is of great important.
One of the most common digital image formats is JPEG,
which is a standard file format for compressing pictures.
Copy-paste tamper between JPEG images is a common
tampering situation, which will involve JPEG compression.
Fridich[1] et al first proposed the method for estimating the
original quantization step of double compressed JPEG.
Luo[2] et al researched a method for detecting cropped and
recompressed image block. And Li[3]et al utilized the
mismatch information of block artifact grid as a clue of
copy-paste forgery. Meng[4] et al proposed an approach to
detect image tampering based on the statistical
characteristics caused by double JPEG compression, but it is
only valid for the uncropped image. A new method of
detecting clipped double JPEG compression image is
presented in this paper, which is based on the method of
Meng and makes use of the theory of the blocking effect
caused by double JPEG compression proposed in paper [5].
The method is available to both cropped and uncropped
double compressed image.
II. JPEG
GHOST
Farid[6,7] et al proposed that there exists two closely
related quality factors of the double compressed image if it
is compressed by them. For example, if a picture is
compressed by quantization matrix ܳͳ at the first time and
ܳʹ at the second time, then we use the method proposed by
farid to get the error distribution plot, and it is likely to have
two obvious “break” in the error distribution plot at the
position ܳͳ and ܳʹ .
Take image “lena”(256*256,BMP format) for example.
We get image 1 through quantizing the coefficients by
quality factor ݍͳ =15. And then get image 2 which is
quantized by ݍͳ=15 and ݍʹ=18 in order(the quantization is
defined as 1 to 20, which is corresponding to the quality
factors ranging from 5% to 100%). Utilizing the method
proposed by Farid, we get the error distribution plot that is
presented in Figure 1.
a)The result of image 1 b)the result of image 2
Figure 1. Plot of error Distribution
In figure 1 a), the break appears at ݍ=15, so it is easy to
know the quality factor of image 1 is 75%. In figure 1 b),
the position of break is corresponding to the ݍ=15 which is
JPEG Ghost named by Farid , and there is an obvious break
at ݍ=18,which is corresponding to the quality factor of
image 2 .
III. E
STIMATION FOR THE OFFSET OF NON-ALIGNED GRIDS
In order to carry on JPEG compression to a picture, we
should firstly divide the image into blocks of 8×8, and then
2015 International Conference on Intelligent Information Hiding and Multimedia Signal Processing
978-0-7695-5668-0/15 $31.00 © 2015 IEEE
DOI 10.1109/IIH-MSP.2015.27
185