Abstract:
In order to solve the problem of low computation time and low detection accuracy in the traditional Copy-Move tampering method based on key-points, an image Copy-Move forgery detection scheme is proposed. The proposed method firstly uses SIFT algorithm to extract image basic features, and the discrete wavelet transform (DWT) and gaussian random matrices are used to reduce the dimension of SIFT feature vectors to achieve feature matching. Then, the affine transformation matrix between the copied region and the pasted region are calculated by using the RANSAC algorithm, and the outliers of mismatching is are eliminated. And the end of the article, the MICC-F220 standard database is simulated experimentally. The experimental results indicate that the proposed detection scheme has great improvement in the computation time and detection accuracy and has good robustness to geometric transformations such as translation, rotation and scaling, as well as some content-preserving manipulations including JEPG compression, wiener filtering and adding salt and pepper noise.