Blind Detection of Copy-Move Forgery Based on Multi-Scale Autoconvolution Invariants

With regard to the Copy-Move forgery of image region, this paper proposes one blind detection algorithm based on Multi-Scale Autoconvolution. The eigenvectors of image are generated by extracting the MSA invariants of each image block, and sorted by dictionary ordering. Then the similarity of image blocks are computed by using correlation coefficients in order to detect and locate the tampered image regions. The experimental results show that the algorithm can not only effectively detect and locate the duplication regions, but also resist part of post-processing operations including rotation attack, Gaussian noises attack and JPEG compression attack. It demonstrates that our proposed algorithm has advantages of low time complexity and high robustness.