Passive forensic method for detecting duplicated regions affected by reflection, rotation and scaling

Most of the existing forensic methods aimed at detecting duplicated regions in images are sensitive to geometrical changes in the replicated areas. Consequently, a simple reflection/rotation of the copied region could be used to not only suit the scene of the image, but also hinder its proper detection. In this paper, we propose a novel forensic method to detect duplicated regions, even when the copied portion have experienced reflection, rotation or scaling. To achieve this, overlapping blocks of pixels are re-sampled into log-polar coordinates, and then summed along the angle axis, to obtain a one-dimensional descriptor invariant to reflection and rotation. Moreover, scaling in rectangular coordinates results in a simple translation of the descriptor. This approach allows us to perform an efficient search of similar blocks, by means of the correlation coefficient of its Fourier magnitudes. Results are presented to demonstrate the effectiveness of the proposed method.