IAI CFVE study materials
IAI CFVE study resources

Twenty-Nine Point Nine Seven Is Not Thirty

IAI CFVE · Put your knowledge to work in context.

All case studies

Twenty-Nine Point Nine Seven Is Not Thirty

A bank ATM vestibule camera records 720×480 interlaced video at 29.97 frames per second (59.94 fields). The relevant event is a card skimmer being installed, lasting about 6 seconds. The bank's security vendor supplies a 'court-ready' MP4 that has been deinterlaced to 30p and re-encoded at 2 Mbps. When you step through it, you see combing on motion, occasional doubled limbs, and a frame count of 187 for the 6.2-second clip. The vendor cannot supply the raw export because 'the system only outputs MP4.' You need to identify the exact frame where the suspect's hand leaves the skimmer and to state the time relative to the ATM transaction log, which records 14:32:07. The defense may argue the deinterlacing created or hid frames. The ATM screen is visible in the same camera view and displays the transaction time.

1. Explain why the clip has 187 frames and why combing and doubled limbs appear.

Read the worked answer

At 29.97 frames per second, 6.2 seconds contains about 185.8 frames, so 186 frames is the expected count; 187 can arise from inclusive counting, an extra head or tail frame, or frame-rate conversion rounding. The source is interlaced: each frame contains two fields captured 1/59.94 second apart. If the vendor blended the two fields into one progressive frame, a moving hand appears in two positions-combing and doubled limbs. If the vendor bob-deinterlaced to 59.94p and then conformed to 30p, the frame count and temporal sampling change again. The 2 Mbps re-encode also removes high-frequency detail. None of this proves fabrication, but the output is not frame-accurate: one output frame may represent two instants, and a 'frame' in the MP4 is not necessarily a unique sample of the original.

2. What would you request or do to recover frame-accurate evidence?

Read the worked answer

First request the native export at 59.94 fields, or a lossless export with no frame-rate conversion, plus the vendor's exact deinterlace settings, encoder version, and export log. Ask whether the ATM recorder stores fields separately and whether the MP4 was produced by the recorder or by a separate transcoder. If only the 30p file exists, ask for a version exported at 59.94p or with 'bob' deinterlacing rather than blending, and verify with a test pattern or timecode burn-in. Reproduce the vendor's process on a known reference clip to show how it handles motion. If no better file can be obtained, treat the supplied file as a derivative and avoid claiming field-level precision. Document that the original remains the best evidence and that the vendor's statement about output format should be tested, not accepted.

3. How would you express the frame and time of the hand leaving the skimmer relative to 14:32:07?

Read the worked answer

If the original is available, identify the field and report both field index and time, e.g., field 372 at 6.208 seconds after the clip start, with field parity stated. If only the 30p file exists, report the output frame index and explain that each frame may span two source instants. Use the visible ATM screen to calculate the offset between the clip and the transaction log, then state a range-for example, '14:32:06.8 to 14:32:07.3'-rather than a single frame as ground truth. Include the basis for the offset, the measured drift if any, and the conversion uncertainty. This preserves the defense's ability to test your method and avoids overstating precision that the derivative file cannot support.