Interpolation method identification from resampled images

This paper aims to identify the type of interpolators used for resampling. Mostly three types of interpolators are widely used: nearest neighbor (i.e., order zero), bilinear (i.e., order one), and bicubic (i.e., order three) interpolation methods. The proposed method directly applies the discrete Fourier transform to an image and computes logarithm of the spectrum magnitude. The spectrum shows specific patterns according to the type of interpolators. In addition, direction of interpolation can also be identified. The type identifier can be incorporated with heuristic classifiers for automatic identification.