A Thorough Investigation Into the ENF Reconstruction in Videos Exposed by Rolling Shutter

In electric network frequency (ENF)-based video forensics, the analysis of videos captured by rolling shutter systems, where each row of a frame is exposed at different time instances, is critical. To gain the advantage of increased sampling frequency in these videos, in contrast to those captured by the global shutter where an entire frame is exposed at a time, the ENF-related luminance signal that is essential for ENF estimation is built by concatenating ENF-related luminance estimates across consecutive frames. However, this approach brings about some issues or phenomena owing to an idle period at the end of each frame. First, the ENF harmonics may be replaced by new ENF components and attenuated, thereby affecting the reliability of the ENF estimates from these videos. Another critical phenomenon is ENF reversal, which is yet to receive much research. This study comprehensively investigates this phenomenon to explore how and under what conditions the ENF is reversed. Further investigations led this study to examine how the ENF in the emerging components is mainly reconstructed from multiple ENF-related luminance harmonics, depending on the idle period. This helps identify reliable ENF components from which the ENF signal can be accurately estimated. In addition, it reveals the optimal idle periods for any ENF component. Using this outcome, this study also proposes a technique to enhance the effectiveness of an ENF component based on idle period modification. The experimental results show that the proposed method may boost the efficiency of an unreliable ENF component, outperforming the existing techniques.