基于噪声水平估计的帧复制篡改取证算法

基于视频噪声的时域变化规律,提出一种可鉴别帧复制篡改的噪声水平检测方法。对各视频帧实施小波变换,利用小波系数的绝对中位差估计各视频帧中混入高斯噪声的标准差,并对标准差时域序列进行快速傅立叶变换,计算幅频谱的峰均比,再通过对峰均比作硬阈值判决,判断标准差时域序列是否存在周期性,自动识别帧复制篡改。结果表明,噪声水平检测方法可确保伪造视频幅频谱具有较大的峰均比,检测准确度比较高,相比于现有检测方法,避免噪声干扰带来的性能损失,表现出较好的检测性能。

Detecting frame of repetition forgery based on noise level estimation

Detecting method of the varying noise level in temporal-domain was investigated based on noise-level, which could identify frame repetition (FR) forgery. Wavelet coefficients were computed for each video frame, and median absolute deviation (MAD) of wavelet coefficients was used to estimate the standard deviation of Gaussian noise mixed in each video frame. Fast Fourier transform (FFT) was used to calculate the amplitude spectrum of the standard deviation curve of the video sequence, and to provide the peak-mean ratio (PMR) of the amplitude spectrum. In order to automatically identify FR forgery, a hard threshold decision based on PMR was taken to determine whether the standard deviation had a periodicity in time domain. The experimental results showed that the proposed method ensured a large PMR for the forged video and high detection accuracy. The proposed method presented a better detection performance when compared with the existing detection, avoiding the performance loss from noise.