Dyadic wavelets and DCT based blind copy-move image forgery detection

This paper proposes a blind method of copy move image forgery detection using dyadic wavelet transform (DyWT) and discrete cosine transform (DCT). An input image is decomposed using DyWT to approximation (LL) subbands and detail (HH) subbands. DCT is then applied to overlapping blocks in LL and HH subbands, and Euclidean distances between the blocks are calculated using DCT coefficients. Decision is made based on similarity of the blocks in LL subband and dissimilarity of the blocks in HH subband. The proposed method is evaluated with images of different sizes, different compression qualities, and with or without rotation before pasting. Experimental results show that the method performs better in all cases than two other multiresolution based methods.