High precision rotation angle estimation for rotated images

The detection of image rotation and the estimation of rotation angle are of great help to “copy-paste” image forgery detection because the copy-pasted slice is often rotated to gain better visual effect. This paper proposes a novel blind image rotation detection algorithm with high precision rotation angle estimation. The proposed algorithm is based on the periodicity within pixel variances, which is introduced by the rotation interpolation. Experiment results show that this algorithm can estimate rotation angle precisely with a fine rotation angle resolution of 0.2 degrees for images of 1024 pixels in width.