Abstract
The number of forged images is currently expanding vastly over the Internet. Therefore, image authenticity represents a globally challenging issue that must be addressed. Emerging tools for image editing have been developed to manipulate and enhance digital images; however, forgers can exploit these tools to achieve their destructive purposes. Forgers often use a common method of image forgery called region duplication forgery. In this method, the copied region in the fake image can appear identical to the original region of the image. This paper aims to target this issue by developing an algorithm that can detect suspected images through localizing small duplicated regions. These regions can be described by multiscale features, which are invariant with illumination variations. The proposed method begins with segmenting suspected images using an adaptive statistical region merging. The goal of the segmentation method is to discover small regions. The method then targets the small regions based on color correction to represent their illumination features. Experiments are also conducted to validate the proposed method on two image datasets, Media Integration and Communication Center (MICC) and Image Data Manipulation, yielding positive results. A comparative study of the most recent methods is carried out.