- 期刊
FI-SIFT Algorithm for Exposing Image Copy-Move Forgery with Reflection Attacks
摘要
In order to improve the robustness of SIFT algorithm to reflection attack, a flip-invariant SIFT (FI-SIFT) descriptor is proposed to detect copy-move forgery of digital images based on the study on the arrangement of SIFT descriptor after reflection attack in this paper. The pro- posed descriptor FI-SIFT is designed to improve the in- variance to reflection and perform as well as SIFT in other situations. Our method starts by extracting FI-SIFT descriptors for detected SIFT key points in the suspicious image. Then, the g2NN method is adopted to implement multiple key points matching. Next, the possible affine transform between matched key points is estimated to re- move the mismatched key points. Extensive experimental results are presented to confirm that our method performs well to detect copy-move forgeries distorted by common attacks including rotation, scaling, reflections and their mixture, especially for the sophisticated scenario, such as multi-objects forgery with combination of reflections.