Tampering detection using hybrid local and global features in wavelet-transformed space with digital images

Abstract

The widespread availability of advanced digital image technology and powerful image editing tools has made it extremely easy to manipulate image content. One popular technique that is challenging for tampering detection methods is the copy–move forgery. Here, one part of the image is copied and is pasted into another part of the same image. Image tampering is very difficult to detect with the naked eye. The authenticity of digital images is critical when they are used as evidence in court, for news reports or insurance claims, as they have the power to influence decisions and outcomes. Hence, this paper presents an efficient method for copy–move forgery detection by means of a HOG descriptor and local binary pattern variance algorithms. The copy–move forgery detection (CMFD) approach is introduced and is applied to the forged region by determining suitable post-processing techniques. The proposed technique is evaluated using MICC-F220, MICC-F2000, UCID, CoMoFoD and CASIA TIDE data sets in which translation, flipping, rotation, scaling, color reduction, brightness change and JPEG compression are applied to an image. The experimental performance of the proposed technique is assessed in terms of the true- and false-detection rates. Ultimately, our proposed method is highly suitable for detecting the altered region, and accurate CMFD results are obtained in forensic image applications.