You are here

Home » Content

G-SURF ve AKAZE tabanlı yeni bir kopyala-yapıştır sahteciliği tespit yöntemi

A new copy-move forgery detection method based on AKAZE and GSURF

Journal Name:

  • Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi

Publication Year:

  • 2017

Key Words:

Keywords (Original Language):

Author NameUniversity of Author
Abstract (2. Language): 
Copy move forgery, one of the most common type of image forgeries, copies a portion on the image and pastes it onto another region on the same image. Main purposes of this forgery type are either replicate or conceal an object in the image. Copy move forgery detection methods can be classified into two groups: Block based and Keypoint based methods. Keypoint based methods have gained popularity lately because they can detect forged regions faster. Two of the most important problems of these techniques are, linear scale space used in creating scale images and blurring whole image without preserving edges, which feature extraction methods rely on. In this work, we used both AKAZE keypoint extraction method, utilizing non-linear scale space to construct scale images and G-SURF descriptor extraction algorithm, relying edge information during the descriptor construction. The method also uses RANSAC algorithm to eliminate false matches. AKAZE is a keypoint extraction and descriptor construction algorithm using non-linear scale space. Keypoint extraction methods in the literature construct scale space of an image by using approximation of Gaussian. Main disadvantage of these methods is they do not respect the natural boundaries of objects in the image. These algorithms apply smoothing operation on image diminishing both noise and edge information at the same time. Thus keypoint information on the edge regions of the image cannot be preserved by these methods. However, AKAZE constructs non-linear scale space of the image and applies adaptive blurring operation on image. G-SURF descriptor extraction algorithm use second order multi-scale gauge derivatives to construct descriptors and it also utilizes per pixel information to make blurring adaptive according to the content of the image. In this work we used Gauge-SURF descriptor with 20s×20s square grid. RANSAC is also used by the method to eliminate false matches. This algorithm determines a randomly created set from matched keypoints and it constructs a transformation matrix by using this set. This transformation matrix evaluates other matched keypoints and some of them are indicated as outlier. This procedure can be applied number of times to approximate real solution. Proposed method extracts keypoints from the test image using AKAZE keypoint extraction algorithm and it then constructs descriptors for each keypoint using G-SURF descriptor extraction algorithm. Matching keypoints are determined after descriptors are extracted. The method uses k-nn to determine the best match for each keypoint. RANSAC algorithm is applied on the matched keypoints to eliminate false matches as the last step. MICC-F220 database is used to evaluate and to compare the results of the method. The database consists of 110 forged and 110 original images. Experiments indicate that the method has improved True Positive Rate (TPR) with reduced False Positive Rate (FPR) compared to similar works reported in the literature. The method gives better results when both TPR and FPR are considered together. The results also indicate that the method has higher recall rates under various attacks. Precision values are also better than the similar works when JPEG compression and Gaussian blurring attacks are considered. However, the method gives worse results when scaling attack is considered. AKAZE detects less keypoints on the pasted regions compared to SIFT method when rescaling attack is applied before pasting operation. Thus, the method cannot detect some keypoints on the forged regions when compared to the method in (Amerini et al., 2011). Usage of Gauge-SURF descriptor becomes meaningless due to the artefacts of keypoint extraction method. We plan to improve the method with segmentation algorithms in the future. Test image can be segmented into regions depending on colour or pattern information and similarity can be tested among regions to improve the precision of the method. We also aim to improve AKAZE keypoint extraction method to make it more robust against scaling attacks. Blurring parameter, which is used during scale space construction, can be adaptively chosen according to the characteristics of the current test image.
Bookmark/Search this post with
Abstract (Original Language): 
Görüntü sahteciliği yöntemlerinden sıklıkla karşılaşanı olan Kopyala-Yapıştır Sahteciliği, görüntü içerisindeki bir bölgenin başka bir bölgenin üzerine kopyalanması ile gerçekleştirilir. Bir nesnenin kapatılması yada var olan bir nesnenin tekrarlanması bu tür sahteciliğin amaçları arasında yer almaktadır. Kopyala yapıştır sahteciliğini tespit etmeye çalışan yöntemler ikiye ayrılmaktadır: Blok tabanlı ve Anahtar noktası tabanlı yöntemler. Son yıllarda ise Anahtar noktası tabanlı yöntemler daha hızlı bir şekilde sahte bölgeyi tespit edebildikleri için araştırmacılar arasında popülerlik kazanmıştır. Fakat anahtar noktası tabanlı yöntemlerin en önemli problemlerinden biri kullandıkları ölçek uzayının lineer olması ve ölçek uzayı görüntülerini oluştururken görüntünün her bölgesine eşit bulanıklaştırma uygulamasıdır. Çalışmada doğrusal olmayan ölçek uzayını ölçek görüntülerini oluşturmada kullanan AKAZE (Accelerated KAZE) anahtar noktası çıkarma yöntemini, yine tanımlayıcı elde etme aşamasında görüntünün kenar bilgisini koruyan G-SURF (Gauge-Speeded-Up Robust Features) tanımlayıcı elde etme algoritması ile beraber kullanılmıştır. Aynı zamanda hatalı eşleşmeleri ortadan kaldırabilmek amacı ile RANSAC (Random Sample Consensus) algoritmasından faydalanılmıştır. Önerilen yöntemin sonuçlarını değerlendirebilmek ve kıyaslama yapabilmek amacı ile MICC-F220 veritabanı kullanılmıştır. Deneyler yöntemin Doğru Pozitif Oranı ve Yanlış Pozitif oranı açısından benzer yöntemlerle kıyaslayınca daha iyi sonuçlar ürettiğini göstermiştir. Sonuçlar aynı zamanda yöntemin çeşitli saldırılar karşısında daha yüksek Seçicilik değerleri elde ettiğini ortaya koymaktadır. Duyarlık değerleri ise özellikle JPEG sıkıştırma ve Gauss bulanıklaştırma saldırıları değerlendirildiğinde benzer çalışmalardan daha iyi sonuçlar üretmektedir.
FULL TEXT (PDF): 
PDF icon arastirmax-g-surf-akaze-tabanli-yeni-bir-kopyala-yapistir-sahteciligi-tespit-yontemi.pdf
  • 4
781
795
  • Turkish

REFERENCES

References: 

Alcantarilla, P. F., Bartoli, A., Davison, A. J. , (2012).
KAZE Features, European Conference on Computer
Vision (ECCV), Fiorenze, Italya.
Alcantarilla, P. F., Nuevo, J., Bartoli, A., (2013a). Fast
Explicit Diffusion for Accelerated Features in
Nonlinear Scale Spaces, British Machine Vision
Conference (BMVC), Bristol, İngiltere.
Alcantarilla, P. F., Bergasa, L. M., Davison, A. J.,
(2013b). Gauge-SURF descriptors, Image and Vision
Computing, 31, 103-116.
Amerini, I., Ballan, L., Caldelli, R., Bimbo, A. Del.,
Serra, G., (2011). A sift-based forensic method for
copy–move attack detection and transformation
recovery”, IEEE Ttans. Inf. Forensics Secur., 6, 1099–
1110.
Amerini, I., Ballan, L., Caldelli, R., Bimbo, A. Del. ,
Tongo, L. Del, Serra, G., (2013). Copy- move forgery
detection and localization by means of robust
clustering with J-linkage, Signal Process. Image
Commun., 28, 659–669.
Bayram, S., Sencar, H. T., Memon, N., (2009). An
efficient and robust method for detecting copy-move
forgery, Proceedings of IEEE International
Conference on Acoustics, Speech and Signal
Processing, 1053–1056.
Christlein, V., Riess, C., Jordan, J., Riess, C.,
Angelopoulou, E., (2012). An evaluation of popular
copy-move forgery detection approaches, IEEE TIFS,
7(6), 2012.
Cozzolino, D., Poggi, G., Verdoliva, L., (2015). Efficient
dense-field copy- move forgery detection, IEEE TIFS,
10(11), 2015.
Farukh, M., Anand, V., Keskar, A. G., (2014). Copymove
Image Forgery Detection Using an Efficient and
Robust Method Combining Un-decimated Wavelet
Transform and Scale Invariant Feature Transform”,
AASRI Procedia, 9, 84–91.
Fischler, M.A., Bolles, R.C., (1981). Random sample
consensus: a paradigm for model fitting with
applications to image analysis and automated
cartography, Commun ACM, 24, 381-395.
Fridrich, J., Soukal, D., Luka ́s, J., (2003). Detection of
copy-move forgery in digital images, Proceedings of
DFRWS 2003, Cleveland, OH, USA.
Tablo 2. Yöntemlerin tespit doğruluğu açısından
kıyaslaması
Yöntem DPO (%) YPO(%)
(Fridrich vd., 2003) 89 84
(Popescu ve Farid, 2004) 87 86
(Pan ve Lyu, 2010) 89.96 1.25
(Amerini vd., 2011) 100 8
(Li vd., 2014) 70.91 17.27
(Cozzolino vd., 2015) 84.55 17.27
(Mishra vd., 2013) 73.64 3.64
Yöntem 95.2 2.8
794
G. Ulutaş
Hu, H., Zhang, Y., Shao, C., Ju, Q., (2014). Orthogonal
moments based on exponent functions: exponent-
Fourier moments, Pattern Recognit, 47, 2596–2606.
İmamoğlu, M. B., Ulutas, G., Ulutas, M., (2013).
Detection of Copy-Move Forgery Using Krawtchouk
Moment, in: 8th International Conference on
Electrical and Electronics Engineering (ELECO),
311–314.
Ketenci, S., Ulutas, G., (2013). Copy-move forgery
detection in images via 2D-Fourier Transform, 36th
Int. Conf. Telecommun. Signal Process, 813-816.
Li, L., Li, S., Wang, J., (2012). Copy-move forgery
detection based on PHT, Proceeding World Congr.
Inf. Commun. Technol, WICT, 1061–1065.
Li, J., Li, X., Yang, B., Sun, X., (2014). Segmentationbased
Image Copy-move Forgery Detection Scheme.
IEEE Trans. Inf. Forensics Secur. 6013, 1–12.
Lowe, D. G., (2004). Distinctive image features from
scale-invariant key- points, Int. J. Comput. Vision,
60(2), 91–110.
Mishra, P., Mishra, N., Sharma, S., Patel, R., (2013).
Region duplication forgery detection technique based
on SURF and HAC, Sci. World J., Article ID 267691,
1-8.
Pan, X., Lyu, S., (2010). Region duplication detection
using image feature matching, IEEE Trans. Inf.
Forensics Secur, 5, 857-867.
Popescu, A.C., Farid, H., (2004). Exposing digital
forgeries by detecting duplicated image
regions,TR2004–515, 1–11, 2004.
Ryu, S.-J., Lee, M.-J., Lee, H.-K., (2010). Detection of
copy-rotate-move forgery using Zernike moments,
Proceedings of the 12th international conference on
information hiding, 51-65, Canada.
Wang, X.-Y., Li, S., Liu, Y.-N., Niu, Y., Yang, H.-Y.,
Zhou, Z., (2016). A new keypoint-based copy-move
forgery detection for small smooth regions, Multimed
Tools Appl, 1-16.
Weickert, J., Romeny, B.H., Viergever, M. A., (1998).
Efficient and reliable schemes for nonlinear diffusion
filtering, IEEE Trans. Image Processing, 7(3), 398–
410.
Wu, Q., Wang, S., Zhang, X., (2010). Detection of image
region-duplication with rotation and scaling tolerance,
Second International Conference ICCCI., 100–108.
Zhao, J., Guo, J., (2013). Passive forensics for copy-move
image forgery using a method based on DCT and
SVD, Forensic Sci. Int., 233, 158–166.

Thank you for copying data from http://www.arastirmax.com