Copy-Move Image Forgery Detection by DyWT and DCT using Euclidean distances

Abstract — In this paper we have proposed a copy move image forgery detection technique using Dyadic Wavelet Transform. The virtue of DyWT is its shift invariant transform. Firstly the input image is decomposed into LL (approximation) and HH (detail) subbands. These subbands are divided into blocks and DCT is applied on them for further processing. In next step similarities are calculated between blocks. For this purpose we use Euclidean distance. Euclidean distance is best here to find similarity and dissimilarity among blocks. Lists of distances are sorted lexicographically for LL and HH subbands in ascending and descending order respectively. Using threshold matched pairs are obtained from the sorted list as copy and moved blocks. The important characteristic is that similarity between the copied and moved blocks from the LL subband should be high, while that from the HH subband should be low due to noise inconsistency in the moved block. Keywords- Image forgeries, Digital forensics, Copy-Mov