Copy-move forgery detection in the presence of similar but genuine objects

Images often contain Similar but Genuine Objects (SGO), such as two beverage bottles, similar windows, etc. This poses a natural but unexplored challenge to existing copy-move forgery detection methods with an assumption that similar regions are always manipulated for forgery purpose. In this work, we investigate the limitations of the existing CMFD methods under the SGO setting, and propose a new one with performance improvement. Our method consists of the following key steps: first, pyramid scale space and orientation assignment are used for feature extraction to ensure scaling and rotation invariance; second, combined features are applied for effective texture description; third, similar features between two points are matched through RANSAC to reduce false matches; last, tampered regions are located and correlation coefficient are computed on those regions. The experimental results indicate that the proposed algorithm is effective in detecting SGO and copy-move forgery, and compares favorably to existing methods. Furthermore, our method exhibits robustness under geometric transformation and certain forms of post-processing.