a novel method for copy move blind forgery detection ... · research scholar, bhagwant university...

A Novel method for Copy Move BlindForgery Detection technique for Digital

Image using DCT

Asif HassanDepartment of Electronics and Communication Engg.

Research Scholar, Bhagwant UniversityRajasthan , India

[email protected]. V.K. Sharma

Department of Electronics and Communication Engg.President/Vice-Chancellor, Bhagwant University

Rajasthan , Indiaviren [email protected]

April 26, 2018

Abstract

Digital images are the integral part of our daily life. Im-ages share information through newspapers, internet, andmagazines. With ease of availability of image processing andediting tools, the image forgery has become common. Copy-move forgery is one of the many types of image forgery. Inthis type of forgery, a region of image is copied and pastedonto another region of the same image. This work proposesa novel method in which the image is divided into severaloverlapping blocks then the absolute value of Discrete co-sine transform (DCT) of each block is calculated to extractfeature of image. Finally the approximate mean value ofeach block is calculated and block matching method is usedto identify the copied region.

1

International Journal of Pure and Applied MathematicsVolume 118 No. 24 2018ISSN: 1314-3395 (on-line version)url: http://www.acadpubl.eu/hub/Special Issue http://www.acadpubl.eu/hub/

Key Words:Digital Images; Copy Move Forgery; Dis-crete Cosine Transform; Blind forgery Detection

1 INTRODUCTION

Digital image forgery detection methods are classified as active ap-proach and passive approach as shown in Figure1. Active approachrequires some prior information of the image for forgery detectionsuch as watermarking or signature. Passive approach does not re-quire any prior information of image hence it is also known as “blindapproach”.

In a Copy-Move forgery, a portion of the image is copied andpasted into another portion of the same image. This is done tohide an object in an image by wrapping it with a section copiedfrom another part of the same image. The human eye cannot easilydetect the suspicious objects as the copied parts come from thesame image, its important properties will match with the rest ofthe image and hence will not be recognized.

Fig. 1: categories of image forgery detection

The availability of image processing platforms, makes it easy tocreate Copy Move forgery. An example of a Copy Move forgery isshown in Figure 2.

2

International Journal of Pure and Applied Mathematics Special Issue

There are many techniques have been proposed by various re-searchers to detect Copy Move Forgeries. Fridrich et al. [1] pro-posed a method in which the image is divided into overlappingblocks and DCT coefficients for feature extraction. Then, the simi-lar blocks are detected and forged regions are found. Popescuet al.[2] proposed a technique for detecting small fixed-size image blocksusing eigenvalues and Eigen vectors of each block and applyinglexicographical sorting, the duplicate regions are automatically de-tected. Kang and Wei [3] proposed the use of SVD for extractingfeature vector and dimension reduction then Lexicographical sort-ing is applied to detect forged regions. This algorithm is robust andefficient. Lin et al. [4] proposed a fast copy-move forgery detectiontechnique for finding features vectors and dimension reduction thenRadix sort is applied on feature vectors to detect forgery. Huanget al. [5] proposed the detection of copy-move forgery using SIFTalgorithm using feature matching. Li et al. [6] proposed a mothodbased on DWT and SVD and applied lexicographical sorting onsingular value vector and the forged region is detected.

2 PROPOSED METHOD

The proposed method is to examine whether the input image con-tains copied regions or not and to identify the region from where itis copied from the same image. This method also tests the robust-ness of the algorithm for JPEG, BMP and TIF format images. Thealgorithm was tested for geometrical transformations like rotation.The proposed method is explained in the following steps. The colorimage or gray image of any pixel value is taken as input for whichcopy-move detection must be performed. The Color image is con-verted from RGB to Grey. Grey image is retained as it is. The

3


input image is segmented into equal size overlapping blocks. Thenumber of blocks depends on of the size of the image. It is calcu-lated using the equation Block size = 2t , Where t = log2 (M ×N) - 12. Here ‘M’ is number of rows and ‘N’ is number of columns.The minimum pixel of image is considered as 128 × 128. Henceminimum value of t is 2.

The number of overlapping blocks

=M ×N

BlockSize(1)

Then calculate the absolute value of DCT for each Block. Whenthe region is copied and rotated, the DCT value of the block willbe negative making it difficult to find the similarity between theblock. When we take the absolute value of DCT, we always end upwith a positive number, which makes it easy to find the similaritybetween the block.

The Discrete Cosine Transform (DCT) helps separate the imageinto spectral sub-bands of differing importance with respect to theimage’s visual quality. The DCT is similar to the Discrete Fouriertransform. It transforms an image from the spatial domain to thefrequency domain

Calculate approximate value of mean of obtained DCT of eachBlock by using the following Formula

(Mean of test block X (1-α) ≤ Mean of all blocks) and

(Mean of test block X (1+α) ≥ Mean of all blocks) (2)

Then mean of each block is matched with all other possibleoverlapping blocks in the same image. If the similarity is foundthen the blocks are highlighted and considered as copied block.

3 SIMULATION AND RESULTS

The tests are carried out on the Mat Lab R2013A, a system with2 GB RAM and 2.90 GHZ processor, the images with the differentpixel values, saved in JPEG format. The images are tested to checkthe computational speed and robustness of the proposed algorithm.To evaluate performance of the copy move image forgery detectionalgorithm, simulation and results are performed on the Datasets:

4


COVERAGE [7] which contains copy-move forged (CMFD) imagesand their originals with similar but genuine objects (SGOs). COV-ERAGE is designed to highlight and address tamper detection am-biguity of popular methods, caused by self-similarity within naturalimages

Figure 3a shows the Input image where the random rectangu-lar region is copied and pasted onto the same image and Figure3b shows the detected where the highlighted region represents thecopied and pasted region. Figure 4a shows the Input image wherethe asymmetrical region is copied and pasted onto the same imageand Figure 4b shows the detected where the highlighted region rep-resents the copied and pasted region. The test is conducted withouthaving prior knowledge (Blind approach) of the images with differ-ent pixel values.

The performance of the Proposed method to detect Copy MoveForgery is measured by True Positive Result (TPR), which correctly

5


identifies the forged image and False Positive Result (FPR), whichfalsely identifies original image as forged image. The results areshown in Table I.

TPR =forgedimagedetected

totalforgedimage× %

FPR =originalimagedetectedforged

totalno.oforiginalimage× %

TABLE I. TPR AND FPR IN PERCENTAGE

The proposed method attained the accuracy of 99% for rect-angular and Asymmetrical region Copy Move Forgery detection ofimages with different pixel size as shown in Table I.

4 Conclusion

The Proposed method is tested with a strong database. Irregularrectangular region Copy Move Forgery using MICC-F220 and forasymmetrical region Copy Move Forgery using CoMoFoD database.It is found that proposed work is robust for irregular rectangularregion and asymmetrical region Copy Move Forgery. This methodis effective for geometric transformations like scaling and rotationof copied region. The forgery is tested by means of DCT featuresand taking mean of DCT of blocks and comparing with other blocksof the same image. It works without any prior knowledge about theimage (Blind approach). The proposed method is faster and moreeffective for any pixel size compared to existing methods.

References

[1] J. Fridrich, D. Soukal, and J. Luk as, Detection of copy moveforgery in digital images, in Proceedings of the Digital ForensicResearch workshop, Aug. 2003, pp. 5 8.

6


[2] A. C. Popescu, and H. Farid, Exposing digital forgeries bydetecting duplicated image regions, Dept. Comput. Sci., Dart-mouth College, Tech. Rep. TR2004-515, 2004.

[3] X. Kang, and S. Wei, Identifying tampered regions using sin-gular value decomposition in digital image forensics, in Inter-national Conference on Computer Science and Software Engi-neering, 2008, Vol. 3.

[4] H.-J. Lin, C.-W. Wang, and Y.-T. Kao, Fast copy-move forgerydetection, in WSEAS Transaction on Signal Processing, 2009,pp. 188-97.

[5] H. Huang, W. Guo, and Y. Zhang, Detection of copy-moveforgery in digital images using SIFT algorithm, in Pacific-AsiaWorkshop on Computational Intelligence and Industrial Ap-plication, Vol. 2, Dec. 2008, pp. 272-6.

[6] G. H. Li, Q. Wu, D. Tu, and S. J. Sun, A sorted neighborhoodapproach for detecting duplicated regions in image forgeriesbased on DWTand SVD, in Proceedings of IEEE InternationalConference on Multimedia and Expo, Beijing, Jul. 2007, pp.1750-3.

[7] B. Wen, Y. Zhu, R. Subramanian, T. Ng, X. Shen, and S. Win-kler, COVERAGE A Novel Database for Copy-move ForgeryDetection, in Proc. IEEE Int. Conf. Image Processing (ICIP),2016.

7


a novel method for copy move blind forgery detection ... · research scholar, bhagwant university...

Documents