fan search for image copy-move forgery-amalta 2014

31
Fan Search for Image Copy-Move Forgery Detection SONDOS M. FADL, NOURA A. SEMARY, AND MOHIY M. HADHOUD FACULTY OF COMPUTERS AND INFORMATION, MENOFIA UNIVERSITY, EGYPT { SONDOS.MAGDY,NOURA.SAMRI,MMHADHOUD @} CI.MENOFIA.EDU.EG

Upload: sondosfadl

Post on 21-Jan-2017

226 views

Category:

Education


3 download

TRANSCRIPT

Page 1: FAN search for image copy-move forgery-amalta 2014

Fan Search for Image Copy-Move Forgery Detection

SONDOS M. FADL, NOURA A. SEMARY, AND MOHIY M. HADHOUDFACULTY OF COMPUTERS AND INFORMATION, MENOFIA UNIVERSITY, EGYPT

{SONDOS.MAGDY,NOURA.SAMRI,MMHADHOUD}@CI.MENOFIA.EDU.EG

Page 2: FAN search for image copy-move forgery-amalta 2014

Agenda

IntroductionRelated worksProposed methodExperiment resultsConclusions and future work

Page 3: FAN search for image copy-move forgery-amalta 2014

Introduction

As image is better than thousands of words, World Wide Web nowadays contains a large amount of digital images used for effective communication process.

Page 4: FAN search for image copy-move forgery-amalta 2014

Introduction It becomes very trivial for

professionals or non-professionals to edit any pre-existing photographs by using freely available commercial image editing tools, such as Photoshop

In September 2010, Egypt’s largest newspaper, the state-run Al-Ahram, showed a forged photo of world leaders walking the red carpet during Middle East peace talks at the White House.

Page 5: FAN search for image copy-move forgery-amalta 2014

Introduction

Image Forgery

Types:

Copy-Move Image Compositing

Image Enhancement

Page 6: FAN search for image copy-move forgery-amalta 2014

Introduction

Image Forgery Detection Techniques:

Active Methods Passive Methods

Page 7: FAN search for image copy-move forgery-amalta 2014

Introduction

CM image tampering is one of the frequently used techniques to hide or manipulate the content of the image. Some parts of an image are copied and pasted on other places of the same image.

The image is since 1930, Stalin air-brushed his enemies out of photographs. Where a commissar was removed from the original image after falling out of favor with Stalin.

Page 8: FAN search for image copy-move forgery-amalta 2014

Introduction

We detect the duplicated regions by Block Matching strategy, where the image is dividing into equal-size overlapped blocks, then each block is matched with all other possible blocks in the same image .

Preparing

Feature Extraction

Matching

Decision

CM forgery detection steps

Page 9: FAN search for image copy-move forgery-amalta 2014

Related works

(Popescu & Farid, 2004 :) used a principal component ( ) analysis PCA on image blocks as features.

(Lin, Wang, Kao, & others, 2009( ,)Tripathi &,2011others :) - , used average intensity of sub blocks where each

- - block has been divided into four equal sized sub blocks to give features of each block .

Page 10: FAN search for image copy-move forgery-amalta 2014

Related works

(Huang, Lu, Sun, & Long, 2011 :) used DCT ( d iscrete ) , Cosine transform as the discriminative features Duplicated image blocks will be neighboring in the sorted list for reducing time.

(Lynch, Shih, & Liao, 2013 :) using average gray value as a feature . , for each block The blocks are sorted based on features then the

, blocks are placed evenly into Groups then Buckets .

Page 11: FAN search for image copy-move forgery-amalta 2014

Proposed method

Input image with size

MxN

Gray scale conversion

Dividing into blocks with size 8

and overlap 1

Sort blocks

Calculate dif between pair of sorted blocks

 

If dif <T ?

Calculate dis

If dis >Nd?

For each block feature extraction

Block's positions are centers

FS is starting

Calculate dif between pair of similar direction neighbours blocks

 

If dif <T ?

Calculate dis

Draw result

End

Initial D=1, MV=0 , flag=1 and

sim=1

If dis >Nd?

Increase D =D +1 and MV=MV +1

If flag=0?

If sim=0?

sim=0

If D>M or D>N?

flag=0

1

1

No

No

No

No

No

No

No

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Page 12: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

1 .Conversion Stage :If the input image is RGB, it converts the image into the corresponding gray scale version below:

I = 0.228R + 0.587G + 0.114B

Page 13: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

2 .Blocking Stage :For an image of size M×N, the image could be divided into small overlapping blocks of b×b pixels resulting in B blocks where:

Page 14: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

3 .Features Extraction :a. DCT transformation is applied on each block .

b. In order to facilitate length reduction step, DCT coefficients have been reshaped to a row vector in zigzag order.

Page 15: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

c. Some coefficients are truncated to reducing the dimension of the features and reducing processing time, then storing in an matrixC .

e. C matrix has been lexicographically sorted by radix sort and it has been saved in matrix .

Page 16: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

3 .Detection of Duplicated Region:I matrix, Assume two adjacent blocks feature vectors and, the difference between pair of sorted blocks is calculated by the following below:

where L is the length of the feature vector .If is less than a threshold T, then two blocks are supposed to be similar.

Page 17: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

Spatial distance between two suspected blocks to eliminate the false positives is tested below :

where ( , ) is the position of block i and ( , ) is the position of block i+1, Consider only

Once a match is detected, the matched blocks are retrieved as C1 and C2, and the position of both block are set to (x1,y1) for C1 and (x2,y2) for C2 .

Page 18: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

Applying Fs method:

Algorithm 1: Fan SearchInput: the two duplicated blocks B1 and B2 Procedure:

.1Initialize two variables Flag=1 and Sim =1, which refers to the possibility of comparison between the blocks.

.2Distance (D) and Move variable (MV) are initialize by D=1, MV=0.

.3Determine the eight neighbours blocks ( and ) for , where:

 

,,

,,

,,

and

 

.4Perform the same procedure for neighbours.

.5Compare between blocks pairs with similar direction.

.6If blocks are similar, set Sim=1and mark the regions in the map image .

.7Increase MV by 1, and goto step 3.

.8If the new eight neighbours are the same eight neighbours examined before, D is increasing by 1, and go to step 3.

.9If D is greater than image size (M or N) in any direction, set Flag = 0 and stop the comparison .

.10If blocks are not similar, set Sim =0 and stop the comparison .

Page 19: FAN search for image copy-move forgery-amalta 2014

Proposed method stages

Example on FS algorithm

Page 20: FAN search for image copy-move forgery-amalta 2014

Experiment results

Experiment method and procedure :The experiments were carried out on the Matlab R2012a, RAM 4 GB and processor 2.30 GHZ .

All the images were 128×128 pixels gray image saved in BMP format .

All the parameter in the experiment were set as: T=0.3 , Nd=16 and L=9 .

Page 21: FAN search for image copy-move forgery-amalta 2014

Visual Result

Experiment 1:Shown are the different possible positions of duplicated regions in (a) horizontal, (b) vertical, (c) diagonal, and (d) anti diagonal directions.

Page 22: FAN search for image copy-move forgery-amalta 2014

Time(s)

Lynch (2013)

Huang (2011)

Tripathi (2011)

FS

7.68 4.7005 6.4018 1.2981shows the performance time of

FS compared to other methods .Note that, the proposed method (FS) decreased the processing time up to 75% faster .

The performance time of different methodsVisual Result

Page 23: FAN search for image copy-move forgery-amalta 2014

  Lynch (2013)

Huang (2011)

Tripathi (2011)

FS

Precision 97% 99% 80% 99%Recall 95% 99% 75% 98%Precision and recall has

been calculated as the metrics for quantifying the accuracy of forgery detection.

The precision and recall for different methods. Visual Result

Page 24: FAN search for image copy-move forgery-amalta 2014

Visual Result

Experiment 2:An irregular region is copied and pasted in a non-overlapping area .

Figure shows result with irregularly regions.

Visual Result

Page 25: FAN search for image copy-move forgery-amalta 2014

The precision for different operations .

Experiment 3:The tampered images were distorted by different processing operations :

 

Precision

BMP JPEG (Q=60) JPEG (Q=90) Blur99% 40% 83% 91%

Visual Result

1. JPEG compression with different quality (Q=90 and Q=70). 2. Gaussian blurring. 3. Rotation with angle (less than 5°). 4. scale with some few factor.

Page 26: FAN search for image copy-move forgery-amalta 2014

Experiment 4:The system has been examined with different values of T . Results of some images with variant values of T; first

row presents the original images (left) and tempered images (right). Second row; White detection (left) refers to T=0.1, Red detection (right) refers to T=0.3. Third row; Green detection (left) refers to T=0.5, Blue detection (right) refers to T=0.7.

Visual Result

Page 27: FAN search for image copy-move forgery-amalta 2014

 Threshold

Detection ratesTrue Positive False Positive

0.1 100 % 0% 0.3 99 % 1 % 0.5 96.5 % 3.5 % 0.7 91.7 % 8.3%

Experiment 4:Table 4 presents the number of true positives and false positives for various T.

The true positive and false positive for different threshold.

Visual Result

Page 28: FAN search for image copy-move forgery-amalta 2014

Conclusions

In this work, we have proposed a fast and efficient method for CM forgery detection, by using fan search method, which compare only neighbors of suspected blocks .

The experiment results show that the proposed method has the ability to detect CM forgery in an image faster than other systems by about 75%.

Page 29: FAN search for image copy-move forgery-amalta 2014

Future work

The method is to be improved for detecting duplicated region under the influence of geometric transformations such as rotation with any angel, scale…etc.

Page 30: FAN search for image copy-move forgery-amalta 2014

References

Khana, A., Malika, S.A., Alib, A., Chamlawia, R., Hussaina, M., Mahmoodc, M.T., Usmand, I.: Intelligent reversible watermarking and authentication: hiding depth map information for 3D cameras. Information Sciences 216 (2012) 155-175.

Hsiao, J., Chen, C., Chien, L., Chen. M.: A new approach to image copy detection based on extended feature sets. IEEE Transactions on  Image Processing, 16.8 (2007) 2069-2079 .

Ling, H., Cheng, H., Ma, Q., Zou, F., Yan, W.: Efficient image copy detection using multiscale fingerprints. IEEE Magazine of Multimedia 19.1 (2012) 60–69 .

Nikolopoulos, S., Zafeiriou, S., Nikolaidis, N., Pitas, I.: Image replica detection system utilizing R-trees and linear discriminant analysis, Pattern Recognition. 43.3 (2010): 636–649 .

Huang, Y., Lu, W., Sun, W., Long, D.: Improved DCT-based detection of copy-move forgery in images." Forensic Science International 206.1 (2011) 178-184.

Popescu, A.C., Farid., H.,: Exposing digital forgeries by detecting duplicated image regions. Dept. Comput. Sci., Dartmouth College, Tech. Rep. TR2004-515 (2004).

Lin, H., Wang, C., Kao, Y. : Fast copy-move forgery detection. WSEAS Transactions on Signal Processing 5.5 (2009) 188-197 .

Page 31: FAN search for image copy-move forgery-amalta 2014

References

Tripathi, R.C., Singh, V.K.: Fast and efficient region duplication detection in digital images using sub-blocking method. International Journal of Advanced Science and Technology 35 (2011) 93-102.

Blelloch, G., Zagha., M.,:Radix sort for vector multiprocessors. Proceedings of the 1991 ACM/IEEE conference on Supercomputing. ACM, (1991) 666-675.

Lynch, G., Shih, F. Y., Liao, H. Y. M.: An efficient expanding block algorithm for image copy-move forgery detection. Information Sciences 239 (2013): 253-265.

Fridrich, J.: Digital image forensics. Signal Processing Magazine, IEEE 26.2 (2009): 26-37.

Ng, T., Hsu, J., Chang, S., Columbia Image Splicing Detection Evaluation Dataset, http://www.ee.columbia.edu/ln/dvmm/downloads/AuthSplicedDataSet/AuthSplicedDataSet.