Image Splicing Forgery Detection Using Local Binary Pattern and Shift Vector

Abstract

Digital images are the simple and fast way of communication. It can convey huge information in short time. But In our now life days various tools and application are available that manipulates the image without leaving any trace of tampering. So we need to design a system to detect the forged image. There are two types of image forgery first one is copy-move and second one is Image splicing. In this paper will base on the splicing image, and the proposed system will utilize Local Binary Pattern and Shift Vector techniques. The required steps to detect the forgery are pre-processing, feature extraction, feature matching.

Full text

International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
Volume 6 Issue 7, July 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Image Splicing Forgery Detection Using Local
Binary Pattern and Shift Vector
Mohanad F. Jwaid Al-Husseinawi1, Husam K. Salih2, Saif A. Salim3, Ahmed Q. Mohammed4
1, 2Maharashtra Institute of Technology, Information Technology, Pune, India
3, 4Dr. DY Patil Institute of Technology, Department of Computer, Pune, India
Abstract: Digital images are the simple and fast way of communication. It can convey huge information in short time. But In our now
life days various tools and application are available that manipulates the image without leaving any trace of tampering. So we need to
design a system to detect the forged image. There are two types of image forgery first one is copy-move and second one is Image
splicing. In this paper will base on the splicing image, and the proposed system will utilize Local Binary Pattern and Shift Vector
techniques. The required steps to detect the forgery are pre-processing, feature extraction, feature matching.
Keywords: Digital image, Splicing image, LBP, Shift Vector
1. Introduction
In our time, digital image plays a significant part in our life.
It being used as a means of pictorial information in everyday
newspapers and magazines as a proof in courts of law, and in
the medical diagnose field [1]. On the other hand, there are
an advanced of great image processing tools, anyone can
easily modification a real picture and generate a fake image.
Photoshop, GIMP, PAINT COREL are an example of such
an editing tools which can do an alteration in digital images
by changing chunks of an image without leaving the special
effects of the modification in the fake image. These changes
cannot be discovered through human eyes. The image fakes
can hide or add a significant piece in an original image to
misguide the court of law [2]. There are two methods of
image forgery, passive and active approach, where passive
method contain image splicing and copy-move image. The
active method contain digital signature and watermarking.
Image splicing is one of the most common kinds of image
tampering [3]. There are many techniques utilized for
discovery image forgery as: discrete wavelet transform,
principle component analysis, discrete cosine transform,
Singular Value Decomposition, Scale Invariant Feature
Transformation and Locally Linear Embedding. In this paper
will base on LBP, shift vector techniques. There is an urgent
need to design an image forgery detection system in various
fields of Forensic investigation, Criminal investigation,
insurance processing, Surveillance systems, Intelligence
services, Medical imaging and Journalism. Next figures will
present image forgery types.
Figure 1: Image forgery types
2. Image Splicing
Image splicing is a common kind to make a tampered image
where a section from one image is copied and pasted into
another image which products composite picture called
spliced image; cut and join two or more snaps of pictures [2].
This type of fake is a challenging issue from tamper detection
point of view. The complicated forgery may include some
post-processing like blurring, JPEG compression, etc. [4].
That performs the forgery detection very hard. Many
researchers tend to discover techniques that discover this
kind of forgery. There are various methods to discover such
an alteration, some of them dependent on the format of the
image Figure 2 shows one of the example of image splicing
process.
Figure 2: Steps to create image splicing
+
Paper ID: ART20175144
406

International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
Volume 6 Issue 7, July 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Figure 3: Example of image splicing
3. Proposed Algorithm
To detect the splicing forgery image, we must to apply four
steps: pre-processing, feature extraction, feature matching
then Post-processing
3.1 Convert RGB to YCbCr
In this step we change the colour of an input image from R,
G, B (red, green, blue) to the YCbCr colour. First thing we
need to know what the meaning of RGB. Image is stored in
the memory as values of pixels, each pixel include three
bytes, and each bytes include eight bits. So the total values is
equal to 16 million colour in each pixel.
Figure 4: RGB colour
While YCbCr Contain the color information and they can be
highly compressed. It is include two parts, Luminance and
Chrominance. This paper based on the Chrominance, because
the human eyes are less sensitive to chrominance than
luminance.
Figure 5: YCbCr colour
To extract YCbCr from R, G, and B, we need to apply:
𝑌=0.299𝑅+0.587𝐺+0.114𝐵 ………… (1)
𝐶𝑟=0.701𝑅−0.587𝐺−0.114𝐵…….….. (2)
𝐶𝑏=−0.299𝑅−0.587𝐺+0.886𝐵……… (3)
Figure 6: Pre-processing
After pre-processing, input image will dividing into
overlapping block.
3.2 Apply LBP
In feature extraction, we utilize Local Binary Pattern. Where
LBP is a powerful feature for texture classification which has
been used widely in this manner. LBP computed by apply:
LBPp.r = )2i ……… (4)
S (Pi-Pc) = … (5)
Where pc is the gray value of the center pixel and pi
represents eight neighbouring pixels. If pi is smaller than pc,
then the binary result of the pixel is set to 0; otherwise, it is
set to 1.
Figure: 7 computing the original LBP code
Paper ID: ART20175144
407

International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
Volume 6 Issue 7, July 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
After feature extraction step, blocks with similar feature
vectors must be identified, and exact copied blocks are
determined based upon some criteria. To implement this
method, the feature vectors are lexicographically sorted and
similar vectors are determined to specify the forgery.
3.3 Apply Shift vector
In this step we match between the points of images to select
if the upload image is original or forgery. Shift vector is
utilize in this step. Let (x1, x2) and (y1, y2) be the locations
of two similar blocks. The shift vector between blocks can be
computed as in:
Sh = (sh1, sh2) = (x1-y1), (x2-y2) ………. (6)
−Sh and Sh, both shift vectors represents same shifting so
they can be normalized. If required, shift vector is multiplied
with −1 such that sh1 ≥ 0. Counter value k of normalized
shift vector increased block pair with same shifting is
detected.
K (Sh1, Sh2) = k (Sh1, Sh2) +1 ……………… (7)
Firstly, counter value is set as zero. When this method
complete, counter value displays frequency of occurrence of
different shift vectors corresponding to matching chunks. A
threshold is set for occurrence of normalized shift. Sh (1), Sh
(2), Sh (n). . If threshold value is great then some matching
blocks will be detected as non-match. For minor value of
threshold too a lot of incorrect matches will happen.
With shift vectors, matched blocks are discovered. Blocks
with similar shifting are characterized with different colour to
display fake section of input duplicate.
Figure 8: Proposed system architecture
4. Experimental Result
The upload image of the proposed system is being taken from
CASIA dataset. CASIA ITDE V1.0, it is collected an image
set containing 1,721 color images of size 384 _ 256 pixels
with JPEG format. It divided these images into two subsets:
authentic set and tampered set. There are 800 images in the
authentic set and 921 images in the tampered set. The
proposed system is based on two parameters:
TPR = ……. (8)
FPR = … (9)
Table 1: Average detection rate
Method TPR % FPR %
Yu-Feng Hsu and Shih-Fu Chang [7] 70 9
X.Pan, and S.Lyu [6] 83 8.8
Thibaut Julliand_, Vincent Nozick † and
Hugues Talbot [8] 84 5.9
The proposed method 89 8
5. Conclusion
The meaning of image forgery detection discussed in this
paper, there are two types of forgery. The common types of
forgery is splicing image. The proposed method started with
divide the chrominance of input image into overlapping
blocks. Then Local Binary Pattern utilized in feature
extraction. In the feature, matching applied shift vector. The
input image is taken from CASIA V1. The accuracy is
achieved to 89 % and the percent of false is decreased to 8%.
References
[1] Bayram S., Avcibas I., Sankur, and B. Memon N.,
“Image manipulation detection,” Journal of Electronic
Imaging – October - December 2006 – Volume 15, Issue
4, 041102 (17 pages), vol. 15(4), 2006.
[2] Sencar H. T. Memon N. Sutcu Y., Coskun B., “Tamper
detection based on regularity of wavelet transform
coefficients,” Proc. ICIP, International Conference on
Image Processing, 2007.
[3] T. J. De Carvalho, C. Riess, E. Angelopoulou, H.
Pedrini, and A. D. R. Rocha, “Exposing digital image
forgeries by illumination color classification,” IEEE
Trans. Inf. Forensics Secur., vol. 8, no. 7, pp. 1182–
1194, 2013
[4] P. Zhang, and X. Kong, “Detecting image tampering
using feature fusion,” In Proc. International Conference
on Availability, Reliability, and Security, ARES, pp.
335–340, 2009.
[5] Y.-F. Hsu and S.-F. Chang, “Detecting image splicing
using geometry invariants and camera characteristics
consistency”, IEEE ICME, July 2006.
[6] X.Pan, and S.Lyu, Member, "Region Duplication
Detection Using Image Feature Matching", IEEE
transactions on information forensics and security, vol.
5, no. 4, december 2010
[7] Yu-Feng Hsu and Shih-Fu Chang, “image splicing
detection using camera response function consistency
and automatic segmentation”, IEEE, ICME 2007
[8] Thibaut Julliand, Vincent Nozick, Hugues Talbot,
“Automated Image Splicing Detection from Noise
Estimation in Raw Images”, 2017
[9] B. Mahdian and S. Saic. Using Noise Inconsistencies for
Blind Image Forensics. Image and Vision Computing,
2009.
Paper ID: ART20175144
408

International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
Volume 6 Issue 7, July 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
[10] K. Francis, S. Gholap, and P. K. Bora, “Illuminant
colour based image forensics using mismatch in human
skin highlights,” 20th Natl. Conf. Commun. NCC 2014,
2014
[11] S. J. Ryu, M. Kirchner, M. J. Lee, and H. K. Lee,
"Rotation Invariant Localization of Duplicated Image
Regions Based on Zernike Moments," Ieee Transactions
on Information Forensics and Security, vol. 8, pp. 1355-
1370, Aug 2013.
[12] J. Wang, G. Liu, Z. Zhang, Y. Dai, and Z. Wang, "Fast
and robust forensics for image region-duplication
forgery," Acta Automatica Sinica, vol. 35, pp. 1488-
1495, 2009.
[13] A. J. Fridrich, B. D. Soukal, and A. J. Lukáš, "Detection
of copy-move forgery in digital images," in in
Proceedings of Digital Forensic Research Workshop,
2003.
[14] Sokolova, M., N. Japkowicz, Beyond Accuracy, F-Score
and ROC: A Family of Discriminant Measures for
Perfor- mance Evaluation, Advances in Arti_cial
Intelligence, 4304, 1015-1021 (2006).
[15] Jaberi, M., Bebis, G., Hussain, M., Muhammad, G., Ac-
Curate and robust localization of duplicated region in
copy-Move image forgery, Machine Vision and
Applications, 25(2), 451-475 (2014).
Author Profile
Mohanad F. Jwaid Al-Husseinawi was born in
Baghdad, Iraq1988. He received his Bachelors in
Engineering of Computer Techniques from Al-
Mammon University Collage, Baghdad, Iraq 2014. He
took his Masters in Engineering of Information Technology from
Maharashtra Institute of Technology (MIT), Pune University, Pune,
India 2017. His areas of interest include Image Processing,
communications.
Husam K. Salih Jubori was born in Baghdad,
Iraq1989. He received his Bachelors in Engineering of
Computer Techniques from Al-Mammon University
Collage, Baghdad, Iraq 2014. He took his Masters in
Engineering of Information Technology from
Maharashtra Institute of Technology (MIT), Pune University, Pune,
India 2017. His areas of interest include Image Processing,
Artificial Intelligence.
Saif A. Salim was born in Baghdad, Iraq1992. He
received his Bachelors in Engineering of Computer
Techniques from Al-Mammon University Collage,
Baghdad, Iraq 2014. He took his Masters in Computer
Engineering from DR.DY Patil Institute of
Technology, Pune University, Pune, India 2017. His
areas of interest include Big Data, Image processing.
Ahmed Q. Mohammed was born in Baghdad,
Iraq1989. He received his Bachelors in Engineering of
Computer Techniques from Al-Mammon University
Collage, Baghdad, Iraq 2013. He took his Masters in
Computer Engineering from DR.DY Patil Institute of
Technology, Pune University, Pune, India 2017. His areas of
interest include Big Data, Image processing.
Paper ID: ART20175144
409