A fast algorithm of image moments in copy-move forgery detection

In copy-move forgery detection, feature extraction is an important step. The image moments (e.g., Zernike moments, radial harmonic Fourier moments, Exponential Fourier moments) are common feature vectors. In view of the excessive length of running time of most existing feature extraction algorithms, a fast algorithm of image moments in copy-move forgery detection was proposed. Integral expression should be discretized and turn integration turned into sum. In the final discretization expression, we can divide it into two parts. One is the fixed section, the other is the grey level of the image. In the classical computing algorithm of image moments, the two parts should be calanlated Num times to get their product if Num image moments are to be obtoined. A fast computing method was put forward to calculate the two parts independently. The fixed parts are calculated just once and the grey level Num times before the product was obtained. This reduced the running time because of the decline of the computing times. If the resolution ratios of the images are invariant and the numbers of the image moments sufficiently large, the fast algorithm of image moments can greatly reduce the running time compared with the classical method.