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

摘要: 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 broad sense. In this paper, we use noise-level estimation expose MCFI operator, exploit periodicity of varying propose an effective automatic detection method. To guarantee high accuracy, high-pass filtering spike enhancement are both employed extract peak outliers in Fourier domain. Depending on these outliers, design criterion credibility value make final decision. The extensive experiments evaluated hundreds sequences with different spatial resolutions two parameter configurations H.264/AVC have shown that validity proposed method, which has better accuracy for method frame repetition.