I have a sensitive shot that needs a very precise 3D track, and it is giving me a hard time. Upon examination I've noticed that it (super 35 full ap) has substantial gate weave visible (the black gate frame is shaking together with the environment).
How is gate weave related to the optical center? My guts tell me that I should first remove the gate weave (to put my optical axis into the middle and stop it from weaving all over) and then do my 3D track, but maybe I am wrong?
Add comment viewed 374 times Latest activity 11 months ago
Hugh Macdonald
[ Editor ]
Here's my take on it:
In the camera, you have the the optical center and the film gate. These are locked together, and won't move relative to each other.
In the scanner, there may be a scanner gate, and then there is also the image center. These two will also be locked to each other.
I don't know scanners all that well, but I'm going to assume for the sake of this answer that the scanner gate, if it exists, is larger than the camera gate.
So, if we can see the camera gate moving on the final scan, then we know that the optical center was also moving. Whether this movement came from the film moving in the camera or the scanner doesn't make a difference.
In an ideal situation, the lens -> digital image transition would be completely locked off. There would be no movement here at all. This happens in digital cameras, but in film there is an intermediate step.
Think of it as having an object with two transformations on. If either or both of those transformations are moving around, the final result will be moving around (unless, but some insane coincidence, they are both moving in opposite directions - but the chance of this happening with something as random as film jitter is negligible)
As Julian mentioned, you can certainly possibly figure out whether the jitter that you're seeing is camera or scanner-based by looking at whether there is motion blur or not - if it was in the camera, there will be motion blur, whereas if it was in the scanner, there won't be (and therefore you can possibly ask them to re-scan it)
Will
I think that the movement of the film that causes gate weave should only occur between exposures. Consequently, this movement of the film should not result in any visible motion blur, and removing it will not make the cg inaccurate. (You'll just need to 2D matchmove the cg to be consistent with the original weaving plate in the final step. There should be no need for motion blur in the 2D transform node.)
My expectation is that you'll be able to get the most accurate 3D solve if you stabilise the film, so that the tracker only has to worry about the motion of the camera itself. I think that heavy gate weave would manifest as a panning movement in the solve, which is not really correct, and probably not what you want.
Sometimes, if it’s really bad, it can be more of a corner-pin wobble, which can be a real pain to get rid of… Seen more on 16mm footage…
Julian Mann
[ Admin ]
I asked a friend who does a lot of match-move:
so .. gate weave - not rolling shutter .. i suppose if you could stabilize it so that weave is removed this may help (in regards to optical center) .. I'd have to think about it a bit to know if this would affect motion blur or not (often removing jitter before a track causes trouble because then your cg does not have motion blur - unless you add it back in 2D using the stabilize/ destabilze node)
dbr
[ Editor ]
With a perfectly locked down camera, in a perfectly locked down scene, the gate-weave would mean a static point in the scene moves around in the plate, so it seems logically to remove this to do the track..
I can’t help thinking, and I can’t quite get my head around this one, that it makes a difference whether the weave was at the camera end or the scanning end… Although I’m starting to disagree with myself here… I think I need to sit down and get my head around what an inconsistency in each would mean…
This is an interesting point – whether it is scanner weave or camera weave. I suspect that the scan is registered and it is camera weave, in which case the gate moves related to the lens and weave needs to be integrated into the solve (!), but for a telecine/scan weave I would have to prestab. Right?
ian_68
Hi Julik
The gate weave would be best removed/reduced by having the footage re-scanned. If this is not possible, then you are left with the solution you have outlined. Hopefully you have everything else you need for the track like camera info and tracking markers etc.
cheers, Ian
julik
[ Editor ]
Ok, since I found the answer myself. If it is CAMERA gate weave, it belongs in the solve and has to be integrated into the solution. If it is a SCANNER or TELECINE weave (more common on 16mm), it needs to be removed prior to tracking for 3D.
If you have both, you are busted but will likely survive by trying to integrate both jitters into your base solve (also less filtering on the image).
I’m going to have to disagree with you here, Julik… I’ll post my own answer in a minute or so…