The principal method used for digital image forgery is Copy-move forgery (CMF). The replicated region may be rotated or flipped to appropriate the scene enhanced for copy-move forgery. In copy–move image forgery, a section from certain image position is replicated and fixed to a different position of the similar image. Typically, post-processing is applied in order to conceal the forgery. The existing forgery detection methods, which generally follow a common procedure especially for copy- move forgery detection are: (1) pre-processing in which forged images are converted to gray space or color space (2) feature extraction in which where features are extracted from different image regions (e.g., overlapped blocks), (3) feature matching, which obtains matched features to determine the original, suspected forgery regions, and (4) post-processing, which discards inconsistently matched pixels or outliers from matched region and only uses the left pixels to obtain the final forgery detected output. As feature extraction majorly affects the accuracy of detection, such methods are generally categorized into three main types named as block-based methods, segmentation-based methods and key point-based methods. In block-based detection methods, the input image is divided into overlapping regular image blocks, and then a descriptor of each block is calculated by various transforms. To extract features that are not affected by normal distortions (e.g., JPEG compression and noise addition) or geometric-distortions (e.g., rotation and scaling), transforms such as the, Principle Component Analysis (PCA) , (PCET) Moment and YCbCr color, Discrete Wavelet Transform (DWT) , Histogram of Orientation Gradient (HOG) ,Discrete Cosine Transform (DCT), Zernike Moment , Krawtchouk Moment , Fourier–Mellin Transform, Signal Value Decomposition (SVD), Polar Cosine Transform (PCT) , and 1-D reflection/rotation-invariant descriptors are applied to blocks to calculate block features [2].

Block Based Forgery Detection, Mean, Variance, DCT