Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 47

1

MAHARASHTRA INSTITUTE OF TECHNOLOGY, PUNE

Department of Information
Technology
Project stage II presentation on :

Image forgery detection using local


binary pattern and transformation
methods
Submitted by
Under the guidance of
Mohanad F. Jwaid
:

Prof. Trupti Baraskar


Exam Seat No. 5126
MIT, Pune
M.E., I.T.
MIT, Pune

08/04/2017
Content
2

• Introduction
• Motivation
• Types of image forgery
• Literature Survey
• Problem Statement
• Goals
• Proposed Methodology
• Analysis of results and implementation
• Conclusion and Future Direction
• Applications
• Papers Publication / Submission
• References

08/04/2017 Image forgery detection using LBP and transformation methods


Introduction
3

 In today’s digital world, digital images are the foremost source of


information and they are the fastest means of information convey
 Digital images are being used in many fields like military, medical
diagnosis, art piece, photography
 The process of creating fake image has been tremendously simple
with the introduction of new and powerful computer graphics editing
software which is freely available as Photoshop, GIMP, and Corel
Paint Shop
 A specific form of image forgery operation called “copy-move” is
considered one of the most difficult problems in the case of forgery
detection

08/04/2017 Image forgery detection using LBP and transformation methods


Motivation
4

 While performing digital image forgery, it is a usual practice to


combine multiple images (some region/regions of the one image is
replaced with the other) for example, when compositing one person’s
head onto another person’s body
 If both of these images were originally of different JPEG
compression quality, then the digital composite may contain a trace of
the original compression qualities
 If both images are of same compression ratio (i.e., the one which is to
be manipulated by using the other), what happens then??

08/04/2017 Image forgery detection using LBP and transformation methods


Types of image forgery
5

Image forgery

Active Passive

Digital Signature Watermarking Copy-move Splicing

08/04/2017 Image forgery detection using LBP and transformation methods


Example of forgery image
6

Image splicing Copy-move image

08/04/2017 Image forgery detection using LBP and transformation methods


Comparison of image forgery types

Image forgery Advantage Disadvantage Author name


type
1)Computational cost is 1)It required some human Harriet Kaur
less. intervention or specially &
Active 2)Simple knowledge equipped cameras. Kamaljit Kaur
about original image is 2)Extra Bandwidth is needed IEEE2016
available. for transmission.

1)Pre existing digital 1)These techniques based on Charmil Nitin


images and data can the assumption that digital Bharti
Passive gain profit using forgeries may not leave visual &
Passive approach. clues that indicate tampering, Purvi Tandel
so they require different IEEE2016
statistics of an image.

08/04/2017 Image forgery detection using LBP and transformation methods


Literature Survey
8

Ref Paper Title Key Techniques Advantage Disadvantage


No.
1 Image Splicing Forgery LBP, Evaluated on Processing Time
Detection using Local Binary DWT, and Two Different is not evaluated.
Pattern and Discrete Wavelet SVM. Datasets and
Transform.IEEE, 2015 having good
efficiency
2 Copy Move Image Forgery Hessian and It is good feature It need long time
Detection Using Hessian and center-symmetric extraction. to detect the
Center Symmetric Local local binary forgery.
Binary Pattern. IEEE, 2015 pattern
3 Passive Detection of Image DCT, Robust and It need long time
Forgery using DCT and Local LBP, and simple method to detect forgery.
Binary Pattern. IEEE, 2016 SVM.

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
9

Ref Paper Title Key Advantage Disadvantage


No. Techniques

4 Image Forgery Detection Adaptive over Working on Less accuracy


Using Adaptive Over- segmentation, both Splicing in splicing
Segmentation and Feature SIFT, Block and Copy- forgery
Point Matching,IEEE, 2015 based. Move Forgery detection.
Images.
5 Blurred Image Splicing Maximum a Having good conflicts
Localization by Exposing posteriori Accuracy resulted
Blur Type irregularity. (MAP), LDA, sometimes
IEEE, 2015 binary while
classifier. classification
process.

08/04/2017 Image forgery detection using LBP and transformation methods


Comparison of techniques
10

Technique Advantages Disadvantages

The principal component 1) Reduced complexity in 1) The covariance matrix is


analysis (PCA) images. difficult to be evaluated in an
2) Reduction of noise. accurate manner.

Discrete Wavelet Transform 1) DWT is detect the noisy 1)Detects many false matches
(DWT) image, retouching images.
2) It captures both frequency,
location information.

08/04/2017 Image forgery detection using on LBP and transformation methods


Cont.….
11

Technique Advantages Disadvantages

Locally Linear Embedding 1) Detects copy move region as  1) It need some times to
(LLE) well as fused edges. detect forgery.
2) Its ability to deal with large
amounts of high dimensional
data.

Scale Invariant Feature 1)Detects JPEG compression, 1)Detects many false matches
Transformation (SIFT) added noise and geometrical
Transformation.
 

08/04/2017 Image forgery detection using LBP and transformation methods


Problem Statement

12

 To check the authenticity of digital images, if it is original or forgery

Goals
 To briefly introduce what is image forgery
 To give overview of different techniques that is used to detect forged
area
 To improve the performance of copy-move forgery
 Investigate the effectiveness of integrating LBP, DWT, PCA and SVM

08/04/2017 Image forgery detection using LBP and transformation methods


System Architecture
13

Upload reference image Upload target image

Pre-processing Pre-processing

Apply DWT Apply DWT

Feature extraction Feature extraction


(LBP) (LBP)

Feature matching (PCA)

Classification (SVM)

Result
08/04/2017 Image forgery detection using LBP and transformation methods
Proposed Methodology

14

Feature extraction

Feature matching

Classification

Feature extraction ( Local Binary Pattern).


Feature matching ( Principle Component Analysis).
Classification ( Support Vector Machine).

08/04/2017 Image forgery detection using LBP and transformation methods


1.Registration and log in

15

 Register as a user
 Name, phone number, email and password are require to register
 Unlimited users
 Type of security

08/04/2017 Image forgery detection using on LBP and transformation methods


2.Pre-processing
16

 R.G.B color

 Each pixel have 3 bytes of data (one for red, one for green and one
for blue)
 1 byte can store a value from 0 to 255
 (28)3 = 16.777.216 colors in one pixel

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
17

 YCbCr
 Y (luminance)

 CbCr (chrominance)

 Contain the color information and they

Can be highly compressed


𝑌=0.299𝑅+0.587𝐺+0.114𝐵
𝐶𝑟=0.701𝑅−0.587𝐺−0.114𝐵
𝐶𝑏=−0.299𝑅−0.587𝐺+0.886 Y (Luminance)

RGB
(Red, Green and Blue)
Cb and Cr
(Chrominance)
08/04/2017 Image forgery detection using LBP and transformation methods
3. Apply DWT
18

 Used for compression


 If an input image size is 2nX2n pixels at level L then after
decomposition at level L+1 its size will be 2n/2X2n/2 pixels.

DWT

 LL: The upper left quadrant


 HL/LH: The lower left and the upper right blocks
 HH: The lower right quadrant

08/04/2017 Image forgery detection using LBP and transformation methods


4.Feature extraction
19

 Local Binary Pattern.


 S (Pi-Pc) =

13 137 55 0 1 0

75 70 80 1 1
LBP
200 45 110 1 0 1

0 1 0 1 1 0 1 1
= 64+16+8+2+1=91
08/04/2017 Image forgery detection using LBP and transformation methods
Cont.…
20

Apply
LBP

126 65 110

90 45 200

80 60 94

08/04/2017 Image forgery detection using LBP and transformation methods


5.Feature matching
21

 The points are extracted in the reference image must be extracted in


the target image
 match between each point in both images
 PCA used as feature matching.
 Determinant (B-λI) = | (B-λI)| = 0
 Only the duplicated region gives the difference in a number of a point
which is extracted

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
22

The given
LBP PCA target image
is forgery

08/04/2017 Image forgery detection using LBP and transformation methods


6.CLASSIFY
23

 Support Vector Machine (SVM) technique


 SVM is a machine learning technique involves training and testing
stage

 g () = T + w0

Z2 > z1

08/04/2017
Cont.…
24

SVM

Image forgery detection using LBP and transformation methods


08/04/2017
Project Design
25

 Hardware Requirements
Processor Intel core i3
RAM 256MB (min)
Hard Disk 20 GB

 Software Requirements
Operating System Windows7, Windows8,
Windows10
Coding language JAVA (JDK 7.0)
IDE Eclipse Luna
Database MySQL 5.0

08/04/2017 Image forgery detection using LBP and transformation methods


Input of the proposed method

26

 The Data Base used in proposed method called Copy Move Forgery
Detection(CoMoFoD)
 contains 260 forgery images
 The maximum size of image in the proposed system is (285X220)
pixels
 The proposed system tested for 40 forged images
 JPEG, PNG and BMP
 CoMoFoD database is available at http://www.vcl.fer.hr/comofod.
08/04/2017 Image forgery detection using LBP and transformation methods
implementation
27

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
28

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
29

08/04/2017 Image forgery detection using LBP and transformation methods


Parameters
30

 Detection accuracy is defined as ratio of number of forged pixels


which are correctly matched to number of actually forged pixels

DA = X 100

 False Match Rate (FMR) is the Parts of images which does not
involve any forgery but still they are displayed as forged parts

FMR = X 100

08/04/2017 Image forgery detection using LBP and transformation methods


Results
Detection accuracy and false match rate for all input images

31

No. Image size (pixel) Image type DA % FMR%


1 129*93 JPEG 94.214 7.245
2 283*203 JPEG 95.4 4.058
3 227*152 JPEG 94.214 5.654
4 274*182 JPEG 95.896 3.985
5 152*152 JPEG 92.214 7.245
6 152*152 JPEG 91.345 8.324
7 152*152 JPEG 94.125 6.458
8 800*532 JPEG 95.632 3.124
9 201*161 PNG 95 4.25
10 184*138 JPEG 93.552 5.213
11 255*256 JPEG 92 3.254

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
32

No. Image size (pixel) Image type DA % FMR%


12 228*172 JPEG 95.748 3.458
13 252*252 JPEG 96.214 2.145
14 217*161 JPEG 95.124 3.014
15 166*163 JPEG 95.214 2.145
16 200*151 JPEG 94.885 5.021
17 185*131 JPEG 93.201 6.247
18 180*135 JPEG 94 5.354
19 250*248 JPEG 89.214 7.982
20 201*161 JPEG 95 4.651
21 233*141 JPEG 91.5 6.14
22 227*222 PNG 95.222 2.145

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
33

No. Image size (pixel) Image type DA % FMR%


23 152*152 JPEG 94.5 4.124
24 200*123 JPEG 95 4.31
25 220*154 JPEG 95.145 3.786
26 220*123 JPEG 95.22 4.157
27 200*123 JPEG 95.5 3.281
28 250*125 JPEG 93.22 5.995
29 230*129 JPEG 95.654 2.458
30 240*105 JPEG 96 2.514
31 240*135 JPEG 94.021 1.984
32 250*154 JPEG 95.874 3.145
33 250*140 JPEG 94.865 4.120

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
34

No. Image size (pixel) Image type DA % FMR%


34 250*156 JPEG 94.454 4.156
35 250*166 JPEG 95.44 4.545
36 150*150 PNG 90.125 8.145
37 255*253 PNG 96 4.125
38 129*93 BMP 94.6 4.235
39 257*257 BMP 96 2.999
40 220*154 BMP 95 5.145

08/04/2017 Image forgery detection using LBP and transformation methods


Analysis results
35

 In the proposed work there are three groups:


 Small size including (100x90 to 160x165) pixels
 Medium size including (200x160 to 240x180) pixels
 large size (250x200 to 285x256) pixels
 results are achieved by dividing input images into blocks sized (8x8)
pixels
 The time require to detection one image is ~ 30 seconds

08/04/2017 Image forgery detection using LBP and transformation methods


Average value of DA %
36

Size of input image in Type of image Detection accuracy %


pixel
Small size (10 images) JPEG 93.851
100x90 to 160x165 PNG
BMP

Medium size (13 images) JPEG 94.391


200x160 to 240x180 PNG
BMP

Large size (17 images) JPEG 95.091


250x200 to 285x256 PNG
BMP

08/04/2017 Image forgery detection using LBP and transformation methods


Average value of FMR %
37

Size of input image in Type of image False Match Rate %


pixel

Small size (10 images) JPEG 5.149


100x90 to 160x165 PNG
BMP

Medium size (13 images) JPEG 4.609


200x160 to 240x180 PNG
BMP

Large size (17 images) JPEG 3.909


250x200 to 285x256 PNG
BMP

08/04/2017 Image forgery detection using LBP and transformation methods


Cont.…
38

Detection accuracy
95.2
95.091
95

94.8

94.6 Small size Medium size


94.391
94.4
Large size
94.2

94
93.851
93.8

93.6

93.4

93.2 08/04/2017
8x8 block size
Cont.…
39

False Match Rate


5.2 5.149

4.7 4.609

4.2 Small size


3.909

3.7 Medium size

Large size
3.2

2.7

2.2
8x8 block size 08/04/2017
Comparison table of Proposed and other
methods
40

Sr. Method Feature techniques Accuracy %


No.
1 Fahime hakimi, Mahdi Hariri, LBP, DWT and PCA. 95.13
farhad GharehBaghi. IEEE,2015
2 Diaa M. Uliyan, Hamid A. Jalab, Hessian and Center 92
Ainuddin W. Abdul Symmetric LBP
Wahab.IEEE,2015
3 Amani Alahmadi, Muhammad DCT, LBP 97
Hussain, Hatim Aboalsamh.
IEEE,2016
4 Chi-Man Pun, Xiao-Chen Yuan, Adaptive over 95.92
Xiu-Li Bi. IEEE,2016 segmentation and
Feature matching.
5 Khosro Bahrami, Alex C. Kot, Li, Local Blur feature 94.6
and Haoliang Li. IEEE,2016
6 Proposed method DWT, LBP, PCA and 95.091
SVM

08/04/2017 Image forgery detection using LBP and transformation methods


Applications
41

 Forensic investigation
 Criminal investigation
 Insurance processing
 Surveillance systems
 Intelligence services
 Medical imaging
 Journalism

08/04/2017 Image forgery detection using LBP and transformation methods


Conclusion and Future Direction
42

 Image forgery detection problem has been explored


 The proposed system utilize an efficient techniques for detection
forgery based on DWT, LBP, and PCA with SVM
 The use of DWT technique made the process of feature extraction
faster
 according to the result, we can observe the accuracy of the result is
achieved to 95 % and the rate of error is decreased to 3% with the
large size of an input image

08/04/2017 Image forgery detection using LBP and transformation methods


Papers Publication / Submission

43

1. Mohanad F. Jwaid, prof. Trupti N. Baraskar “Study and Analysis of Copy-Move & Splicing
Image Forgery Detection Techniques “, at
IEEE International conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-
SMAC 2017)./ presented.

2. Mohanad F Jwaid, prof. Trupti N. Baraskar “ Detection of copy-move image forgery using Local
Binary Pattern with Discrete wavelet transform and Principle Component Analysis” accepted at 3rd
International Conference On Computing, Communication, Control And Automation 17th - 18th Aug 2017/

accepted in this conference.


3. Mohanad F. Jwaid, prof. Trupti N. Braskar “An Efficient technique for Image Forgery
Detection Using Local Binary Pattern (Hessian and Center Symmetric) and Transformation
Method”, at iPGCon 2017 – 8th Post Graduate Conference for Information Technology/
presented.
08/04/2017 Image forgery detection using LBP and transformation methods
References
44

 [1] Fahime hakimi, Mahdi Hariri, farhad GharehBaghi, “Image Splicing Forgery Detection using Local
binary pattern and Discrete Wavelet transform”, 2015 2nd International Conference on Knowledge-
Based Engineering and Innovation.

 [2] Diaa M. Uliyan, Hamid A. Jalab, Ainuddin W. Abdul Wahab, “Copy Move Image Forgery
Detection Using Hessian and Center Symmetric Local Binary Pattern”, 2015 IEEE Conference on
Open Systems (ICOS), August 24-26, 2015, Melaka, Malaysia

 [3] Amani Alahmadi, Muhammad Hussain, Hatim Aboalsamh, Ghulam Muhamma, George Bebis,
Hassan Mathkour, “Passive Detection of Image Forgery using DCT and Local Binary Pattern”IEEE,
2016
 [4] Chi-Man Pun, Xiao-Chen Yuan, Xiu-Li Bi, “Image Forgery Detection Using Adaptive Over-
Segmentation and Feature Point Matching”, IEEE Transactions on Information Forensics and Security,
2015

08/04/2017 Image forgery detection using LBP and transformation methods


References
45

 [5] Khosro Bahrami, Alex C. Kot, Li, and Haoliang Li, “Blurred Image Splicing Localization by Exposing
Blur Type Inconsistency’, IEEE Transactions on Information Forensics and Security, 2015.
 [6] Mohsen Zandi, Ahmad Mahmoudi-Aznaveh, and Alireza Talebpour, “Iterative Copy-Move Forgery
Detection based on a New Interest Point Detector”, IEEE Transactions on Information Forensics and
Security, 2016.
 [7] Anselmo Ferreira, Siovani C. Felipussi, Carlos Alfaro, “Behavior Knowledge Space-Based Fusion for
Copy-Move Forgery Detection”, TRANSACTIONS ON IMAGE PROCESSING, 2016.
 [8] E. Ardizzone, A. Bruno, and G. Mazzola, “Copy-Move Forgery Detection by Matching Triangles of
Keypoints”, IEEE Transactions on Information Forensics and Security, 2015.
 [9] Rani Susan Oommen, Jayamohan M., Sruthy S., “A Survey of Copy-Move Forgery Detection
Techniques for Digital Images”, International Journal of Innovations in Engineering and technology.
 [10] Charmil Nitin Bharti, Purvi Tandel, “A Survey of Image Forgery Detection Techniques”, IEEE
WiSPNET 2016 conference
08/04/2017 Image forgery detection using LBP and transformation methods
References
46

 [11] Jaberi, M., Bebis, G., Hussain, M., Muhammad, G., “Accurate and robust localization of duplicated
region in copy-move image forgery”, Machine Vision and Applications, 25(2), 451-475 (2014).

08/04/2017 Image forgery detection using LBP and transformation methods


47

Thank You.

08/04/2017 Image forgery detection using LBP and transformation methods

You might also like