Noise-level estimation based detection of motion-compensated frame interpolation in video sequences

Motion-Compensated Frame Interpolation (MCFI) is commonly used to produce the fake high-frame-rate videos, and it can be regarded as a video forgery operation from a broad sense. In this paper, we use the noise-level estimation to expose MCFI operator, and exploit the periodicity of noise-level varying to propose an effective automatic detection method. To guarantee the high detection accuracy, the high-pass filtering and the spike enhancement are both employed to extract the peak outliers in the Fourier domain. Depending on these outliers, we design the criterion of credibility value to make a final decision. The extensive experiments evaluated on hundreds of video sequences with different spatial resolutions and two parameter configurations of H.264/AVC have shown that the validity of the proposed method, which has the better detection accuracy for the MCFI method and the frame repetition.