An Evaluation of Popular Copy-Move Forgery Detection Approaches

Abstract

Abstract-A copy-move forgery is created by copying and pasting content within the same image, and potentially postprocessing it. In recent years, the detection of copy-move forgeries has become one of the most actively researched topics in blind image forensics. A considerable number of different algorithms have been proposed focusing on different types of postprocessed copies. In this paper, we aim to answer which copy-move forgery detection algorithms and processing steps (e. g. , matching, filtering, outlier detection, affine transformation estimation) perform best in various postprocessing scenarios. The focus of our analysis is to evaluate the performance of previously proposed feature sets. We achieve this by casting existing algorithms in a common pipeline. In this paper, we examined the 15 most prominent feature sets. We analyzed the detection performance on a per-image basis and on a per-pixel basis. We created a challenging real-world copy-move dataset, and a software framework for systematic image manipulation. Experiments show, that the keypoint-based features SIFT and SURF, as well as the block-based DCT, DWT, KPCA, PCA and ZERNIKE features perform very well. These feature sets exhibit the best robustness against various noise sources and downsampling, while reliably identifying the copied regions.

Fusion of block and keypoints based approaches for effective copy-move image forgery detection

“…4, the region in the red box is classified as smooth, and we calculate the bound- Step 3. As recommended in Christlein et al (2012), Zernike moments are chosen as block features due to its relatively small memory footprint and high reliability. Zernike moments (Khotanzad and Hong 1990;Ryu et al 2013;Ryu et al 2010;Teh and Chin 1988) are rotation invariant and robust to noise, JPEG compression and even blurred.…”

“…However, this has also caused another issue in term of the reliability and data integrity of digital images. In some application areas, such as news reports, court certification and biometrics imaging, the authenticity of the image is particularly important (Christlein et al 2012;Huynh et al 2015;Sencar and Memon 2008).…”

“…In (Christlein et al 2012), these features are grouped into four classes: frequency domain-based, dimensionality reduction-based, intensity-based, and moment-based features. Feature transformed approaches include block-based discrete cosine transform (DCT) (Fridrich et al 2003), principal component analysis (PCA) (Popescu and Farid 2004), discrete wavelet transform (DWT) (Bashar et al 2010), Dyadic Wavelet Transform (DyWT) (Muhammad et al 2011), and Fourier-MellinTransform (FMT) (Bayram et al 2009).…”

“…In Wang et al (2009) and Ryu et al (2010), the first four Hu moments (HU) and rotation invariant Zernike moments are respectively used. After examining and assessing 15 most prominent feature sets, Christlein et al (2012) recommended Zernike moments as block features due to its relatively small memory footprint and robustness to transformation.…”

“…To this end, fusion of block-based method and keypoint-based method to respectively deal with smooth regions and keypoints regions becomes a natural choice in this context. After examining 15 most prominent feature sets, Christlein et al (2012) recommended Zernike moments as the best block features due to its relatively small memory footprint and high reliability. Moreover, their experimental results also showed that SIFT produced the best results among those keypoint-based methods.…”