Motion Vector Consistency
Definition
In unmodified video, motion vectors across consecutive frames track the same objects and follow smooth trajectories. A deleted or inserted frame breaks the motion field: vectors jump discontinuously in direction or magnitude at the tamper boundary.
- Field
- Digital video forensics
- Detects
- Frame deletion or insertion
- Signal
- Discontinuity in motion vector direction/magnitude
- Data source
- Codec-level motion estimation between frames
Common questions
Why does motion vector analysis work when visual inspection does not?+
A skilfully edited cut can look convincing to the eye because adjacent frames still look plausible together, but the underlying motion field encoded by the video codec still shows an abrupt jump at the exact tamper boundary, a discontinuity the codec computed automatically and independently of how convincing the edit looks.
What can cause a false positive in motion vector analysis?+
Genuine scene cuts, rapid camera pans, sudden lighting changes, or a legitimate change in recording conditions can all produce similar motion vector discontinuities, so an examiner must corroborate a suspected tamper point with other evidence such as container timestamps or GOP structure before concluding editing occurred.
How does this technique differ from frame-hash comparison?+
Frame-hash comparison flags frames that are missing or duplicated by comparing frame content directly, while motion vector consistency analyses the encoded motion field between frames, catching subtler edits where frame content alone looks continuous but the underlying object trajectories do not.
Related terms
- Group of Pictures (GOP)
- The periodic structure of I-frames, P-frames, and B-frames in a compressed video stream. Each encoding generation imposes its own GOP structure. When...
- I-Frame (Intra-Coded Frame)
- A video frame that is encoded entirely from its own pixel data, without reference to any other frame. I-frames mark the start...
- Inter-Frame Residual
- The pixel-level difference between a predicted frame and the actual frame after decoding. In unmodified video, residuals at scene cuts are elevated...
- P-Frame (Predictive Frame)
- A frame encoded as a set of changes relative to a preceding reference frame. P-frames are smaller than I-frames because they only...
- Presentation Timestamp (PTS)
- A value stored in the container for each packet indicating when that frame should be displayed. In an unmodified recording, PTS values...