According to the low robustness characteristics of image tampering with evidence of general move-copy algorithm, this paper proposes a new algorithm. Firstly, it uses the color invariant feature to detect feature points, which improves recall forensic results. Secondly, it uses the RGB color channel to expansion Surf descriptor, which increases the robustness of feature vectors. Then, it defines noise sensitivity coefficients to identify and deal with the image noise points and use sequential similarity detection algorithm to improve feature points registration. Finally, it uses Kendall coefficient and the four hypothesis to constrain the forensic results and get the final result. Experimental results show that the new algorithm reflects higher accuracy and strong robustness and it has strong ability of forensics.
