A Review on Automatic Image Forgery Classification Using Advanced Deep Learning Techniques

Abstract

AbstractDigital images are the representation of an event and considered as an evidence for most of the cases and scenarios. Copy-move forgery is a generic sort of forgery method. The technique for recreating one segment or part of the picture inside a similar picture is called as copy-move forgery. An effective and dependable technique has been created by various authors for recognizing these forgeries for restoring the image credibility. Passive approaches of image forgery detection are very hard to achieve. Copy-move, cut-paste, image splicing, image retouching and lightening condition are the examples of independent forgery techniques. Various techniques have been used by various authors like deep learning, convolution neural network, median filtering detection based on CNN, copy-move forgery detection, ringed residual, discrete cosine transform, U-Net, image splicing forgery detection, etc., with good accuracy on publically accessible databases like CASIA, dataset series of MICC, CoMoFoD, BSDS300, etc. In this paper, we have done a critical analysis of these image forgery detection technologies and the dataset available publically. Comparative analysis based on techniques, model, dataset and accuracy has been performed, and they achieve good accuracy as well.KeywordsCNNCopy-move forgeryCut-pasteImage splicing