Copy-move forgery detection using overlapping block approach with fusion of zernike moment and dct transform

Abstract

Image forgery is a kind of manipulation to conceal important and useful information. Between wide ranges of tampering techniques, copy-move forgery which includes copied area from another part of same image, is one of the frequently used techniques. Adding different type of attack such as noise and rotation are often involved to assimilate tampered image to original one. Thus, an improved technique is required to be developed for efficient and robust duplicated-block detection. The aim of this study is to improve detection accuracy under different kind of post-processing like noise and rotation. In order to achieve this goal, a detection method of copy-move forgery using the combination of discrete cosine transform (DCT) and Zernike moments for localization of duplicated regions is proposed. Since the magnitude of Zernike moments is algebraically invariant against rotation and DCT is insensitive to additive noise, the proposed method is therefore introduced by exploiting these advantages. Firstly, image is divided into blocks of 16x16 pixels. Secondly, for each block, both Zernike moments and also DCT coefficients are computed and sorted lexicographically. Thirdly, for each type of feature (Zernike moments and DCT Coefficients), each pair of blocks is compared to find a possible matching according to a predetermined threshold value. The proposed method is evaluated on ten forged images of MICC-F220 standard dataset. Simulation results demonstrate that precision rate of proposed method under rotation and noise attack is above 75% beside, because of upward trend of True Positive, recall rate of proposed method transcends 90%