2

The problem exists when I try to play the original 3.5 gigabyte asf file, and I have converted it to mp4 at least a dozen times, experimenting with various Handbrake settings relating to frame rate (mostly), to see if something might fix the issue and make everything settle into place. No such luck.

I did discover that up-framerating to 120 fps seemed to help a tiny bit, as did allowing for variable frame rate. But the bottom line is that I still have a 45 minute video that's really difficult to watch - particularly when played back at 4x speed, which is my ultimate goal.

So you can see for yourself, I uploaded the first minute or so of the surveillance video here. Unfortunately, playback at 4x only accentuates the stutter / jerkiness. Is the video missing a frame every ~12 frames or so, and skipping the missing frame's "time slot" altogether? (Forgive my inelegant description - this is pretty new to me.)

Is there any hope for salvaging this footage? I was thinking that creating a new video using only certain frames might work (i.e. only odd number frames, or every third frame) but I'm not sure if the frame skipping is as regular as it appears and I know that I don't know this area well. Hence the hopefully smarter move of turning to people who know what they're talking about.

I appreciate any help/guidance y'all can provide with this. Thanks!

Custom Soundtrack
  • 815
  • 1
  • 6
  • 7
  • The MP4 has a cadence of 1 frame followed by 7/8 duplicates. Then every few seconds, a slice of time (~1s) is missing. Is the asf itself processed? – Gyan May 20 '18 at 19:29
  • Thank you for that information. Unfortunately, I don't know whether the asf file is processed. Is there a way for me to tell or would that require asking the source? (The latter would be difficult. I am a journalist and the longer video was filed as evidence relating to a story I am covering.) Stepping through the asf frame-by-frame in VLC reveals that it's as you describe, 1 frame followed by duplicates and then a jump every few seconds. It it reasonable to think that I can turn this into a more normal-looking video by regularly removing certain frames? – Custom Soundtrack May 21 '18 at 05:10
  • Not really. since the time correspondence is destroyed i.e. if frame 1 corresponds to T=0.0s and frame 9 to 0.125s, then removing the duplicates and retiming frame 9 (now #2) to 0.02s will make video look more jerky since the displayed motion is no longer realtime. Also the missing slices will be much closer in the retimed video and accentuate the stutter. – Gyan May 21 '18 at 05:49
  • Two possibilities: 1) the original video is 8-10 fps and it has been resampled to 60 fps by frame duplication. Pointless and wasteful, really. 2) It was high FPS at capture, and during processing, most interim frames were dropped and 60 fps achieved with duplicates of the nearest retained frame. In either case, unless the video with full fluid motion is available, nothing can be done about this stream - there is a synthetic motion interpolation filter but that's imperfect, slow and I don't know if that's ok in a journalism context – Gyan May 21 '18 at 05:49
  • Thank you. I'll use it as is and people will deal with the jumpiness. You are correct that interpolation wouldn't be acceptable as part of a news product, though if it was disclosed and was used only to increase fidelity to the original that might be okay. Academically, for a clip like this one would that involve manually identifying the frames where interpolation needs to happen? I imagine that detecting jumps would be a challenge. – Custom Soundtrack May 21 '18 at 08:57
  • Well, exact duplicates would be removed; retained frames would preserve their timestamps; the filter would replace the empty time slots. For the missing slices, a scene detector would need to run with some fine-tuning since these are jump-cuts not true scene changes. Once detected, all frames starting at the cut would be offset by the estimated missing time. So, the interpolation will fill this in too. But this is academic as the result won't have verisimilitude. – Gyan May 21 '18 at 09:20

0 Answers0