Copy-move forgery detection of digital images using football game optimization

Abstract

ABSTRACT Copy move forgery (CMF) of digital images is the frequently adapted tampering by simply copying a region of the image and pasting it on to the same image using user-friendly image processing tools. Most of the existing key-point (KP) based CMF detection algorithms involve a hefty number of KPs and huge sets of feature descriptors, which increase their computational burden. In addition, the identified KPs may not spread over all regions of the image and the classical clustering techniques may not optimally classify the feature space into cluster space, thereby affecting the accuracy of the results. This paper suggests a new CMF detection method that considers a small number of the strongest KPs, chosen from both Difference of Gaussian (DoG)-based KPs and FAST-corner KPs, evaluates Scale Invariant Features Transform (SIFT) descriptors, applies discrete wavelet transform for dimensionality reduction, and employs football game based optimization (FGBO). The FGBO belongs to the family of meta-heuristic optimization algorithms, and is used for effectively classifying the feature space into cluster space, with a view to overcoming the drawbacks of classical methods. This article exhibits the superior performances of the developed method over existing methods by presenting results on 500 digital images.