Film capture issue - partial film frames on a video frame. Why?

[brassplyer]

When doing a home brewed film capture by pointing the cam at the projected image, there are double images where portions of two different film frames end up on the same video frame.

If the video is capturing at a significantly faster frame rate - 29.97 frames vs 16 or 18 fps, why would the projected film ever have a chance to lay parts of two frames onto one of the video?

[lordsmurf]

Interlacing: http://www.digitalFAQ.com/guides/video/capture-understand-sources.htm

[jagabo]

NTSC Video cameras capture one field every 1/60 second. When that video is digitized pairs of fields are woven together into frames. What is happening is the film frame changed between the time those tow fields were captured. Then the two fields were woven together.

[Gavino]

If the video is capturing at a significantly faster frame rate - 29.97 frames vs 16 or 18 fps, why would the projected film ever have a chance to lay parts of two frames onto one of the video?

I guess you're thinking that since the duration of each video frame (or field) is shorter than one film frame, then only one film frame can appear during that time? But the two are not synchronised, so a change of film frame can easily occur in the middle of a video exposure. (The underlying problem is not really to do with interlacing, a similar thing would happen with a progressive video camera.)

[brassplyer]

Where it came up was in attempting to do deshaking on captured film to see what happened. The deshaking does work after a fashion, but with some issues. Between the smearing from the original photographer's unsteady hand - which I know causes issues with deshaking anyway - and this "between frame" issue, I get these odd artifacts.

Is frame-by-frame film capture the only way to get a capture that's less problematic if you anticipate doing deshaking?

[jagabo]

Try AviSynth's TFM() and TDecimate() to the actual film frame rate.

[brassplyer]

Try AviSynth's TFM() and TDecimate() to the actual film frame rate.

Thanks. Can you summarize how this would help with the issue I'm talking about?

[jagabo]

TFM() will convert those half/half frames to full frames by taking a matching field from an adjacent frame. That will remove the interlacing but leave you with duplicate frames. TDecimate() will remove the duplicate frames (you have to specify the frame rate). Duplicate frames will screw up the deshaker.

You could also use a deinterlacer (like Yadif()) instead of TFM() but that will leave you with more artifacts.

Post a ~10 second sample...

[brassplyer]

Post a ~10 second sample...

Well, here's 7 secs. Total horizontal motion which is where the problem shows up the worst.

http://www.fileden.com/files/2008/3/11/1809238//7sec.avi

[jagabo]

That video has field blending in addition to interlaced frames. Yadif() and SRrestore() will work better than TFM() and TDecimate()

AVISource("7sec.avi")
Yadif(mode=1, order=0)
SRestore(frate=19.98)

Then deshake that.

[brassplyer]

That video has field blending in addition to interlaced frames. Yadif() and SRrestore() will work better than TFM() and TDecimate()

Interesting, the format you uploaded wont load in a player that does frame advance but playing it in WMP and stopping it randomly, I can' find any "ghost frames" other than the very first frame.

-a little later-

I found one download for Yadif, but SRrestore doesn't even come up on the Avisynth Wiki.

[jagabo]

Yes, after srestore there are no ghost frames (or very few ghost frames with very little ghosting).

The file is simply Xvid video in an AVI. You probably don't have an Xvid decoder installed. The player that is playing the file (but doesn't have frame by frame advance) probably has an Xvid decoder built in.

Yadif: http://avisynth.org.ru/yadif/yadif.html
SRestore: http://avisynth.org/mediawiki/Srestore

Note SRestore requires several other filters (links at the wiki).

[brassplyer]

Yes, after srestore there are no ghost frames (or very few ghost frames with very little ghosting).

The file is simply Xvid video in an AVI. You probably don't have an Xvid decoder installed. The player that is playing the file (but doesn't have frame by frame advance) probably has an Xvid decoder built in.

How did you decide on a 19.98 frame rate?

[jagabo]

I stepped through the video field by field and saw that there were usually three duplicate fields in a row. 59.94 / 3 = 19.98. After conversion with SRestore the motion looked pretty smooth. There does appear to be one missing frame in the middle (between frames 69 and 70) of the pan so maybe the rate should be a little higher.

[brassplyer]

I stepped through the video field by field and saw that there were usually three duplicate fields in a row. 59.94 / 3 = 19.98. After conversion with SRestore the motion looked pretty smooth. There does appear to be one missing frame in the middle (between frames 69 and 70) of the pan so maybe the rate should be a little higher.

Okay.

The basic "look" of the video doesn't seem any different to me from the original DV avi file but I assume without the ghost frames it will lend itself to deshaking better without the weird "flashing" artifacts.