Blind Identification of Image Copy-paste Tampering Based on Logarithm Polar Coordinate Transformation

摘要

To detect image copy-paste tampering, an improved blind identification was proposed in this study. Firstly, according to gray-level and gray-structure, every couple of doubtful blocks can be located. Secondly, for each located couple of doubtful blocks, use logarithm polar coordinate transformation to translate the rotation and scaling in Cartesian coordinate system into translation in logarithm polar coordinate system; and use phase correlation method to obtain rotation angle, scaling size and position offset; and then, the tampered regions would be located roughly. Finally, use geometry method to accurately calculate the rotation angle, scaling size and position offset between the copy-paste regions, and according to these values, the tampered region can be located accurately. Experimental results showed that our algorithm can detect the copy-paste tampering, even with rotation and scaling transformation, and it can resist some post-processing, such as blurring, adding noise, and so on.