sent on october 10, 2025
As you likely know, our job is much more difficult when we’re tasked with performing a video-of-a-video analysis. I’m still batting ~500 in my practice, where half the time I get the native video and half the time I get a video of a video.
It's challenging in many regards: the timestamp might be cut off, video quality suffers, the original metadata is gone, the video is jumpy, etc. Sampling rate can also complicate things.
For instance, what if the original video is being played back at an average of 30 fps with every frame being displayed for 1/30 s? If the iPhone is recording at 30 fps with a shutter speed of 1/60 s (a common ratio, but shutter speed varies), you might get lucky with synchronicity between videos and receive clean versions of all original frames. That would look like this (where the top represents the recording and the bottom represents the playback):
However, with a slight misalignment, every iPhone frame is going to be stuck between two original frames. That would result in moving objects presenting in two locations, and conceptually, look like this:
So, what average frame rate would be required to get a clean copy of every original frame? 60 fps, or twice the original frame rate. That’s confirmed visually below; no matter how you offset the green bars, you’ll always get a clean copy of every original frame, often two.
This is a practical application of the Nyquist-Shannon sampling theorem, which essentially states a signal must be sampled at twice its frequency to reconstruct it accurately.
Video-of-video analyses are no walk in the park, and can induce deep, uncomfortable thoughts. However, if you understand the limitations and how to deal with them, they can usually be relied upon for speed analyses, IMHO.
This is a recent thought experiment I’ve been playing with (and graphics), so if you think I got anything wrong, please let me know. Thanks for reading, keep learning!
Lou Peck
Lightpoint | JS Forensics
P.S. If you'd like to learn more about the photogrammetry aspect of video analysis, we just launched a December course!