Stabilizing a jerky telecine

[binba]

I have jerky telecined material (16mm), caused by the use of tape splices on the film. So the jerkiness is very distinct: in the vicinity of every cut ("scene change") the whole frame pans or tilts slightly but annoyingly, 4 to 8 frames long. This is not camera shake, but the frame moving inside the film scanner that was used.

I tried to use DePanStabilize with and without MVtools` vectors, with various settings - with no success. Nothing really got rid of this rather orderly shake.

Two examples, if you have the bandwith and/or patience: (full-frame 24fps DV AVI)
AQ-1 (http://www.sfu.ca/~dlahat/Video/AQ-1.avi) 5MB (watch for the "hiccup" in frames 20-27)
AQ-2 (http://www.sfu.ca/~dlahat/Video/AQ-2.avi) 8MB (frames 50-57)


Excerpts from the scripts I tried:

mdata = DePanEstimate(clip)
DePanStabilize(clip, data=mdata,cutoff=0.4,dxmax=20,dymax=20,info=true)

or

vectors= clip.MVanalyse(lambda=2000,delta=2)
globalmotion = clip.MVDepan(vectors,zoom=false,rot=false,pixaspect=0.911)
DePanStabilize(clip, data=globalmotion,dxmax=10,dymax=10, pixaspect=0.911,info=true)


Any suggestions?

Thanks,
Drew.

[mg262]

The VirtualDub plug-in Deshaker is an alternative to try. (I believe you can call it from AVISynth, but it's probably more hassle than it's worth.)

[rfmmars]

This can't be solved by software. It isn't video frame related. The video frame is smeared from the splice.

You must remove the old splices and make good new ones.

If the film has damaged sproket holes, then only a video printer will work.

Search for Video Fred on the matter.

richard
photorecall.net

[binba]

This is video related, what smear are you talking about? the frame is slightly blurry where the splices are, but this has nothing to do with stabilizing. (From looking at the vectors it doesn't seem to confuse MVanalyse either.) Each splice is 4 frames long, 2 in each shot, the jerks are longer and earlier/later on.

I'll try Deshaker.

And what is video fred?

[mg262]

And what is video fred?
The user @videofred!

Edit - my mistake.

[rfmmars]

This is video related, what smear are you talking about? the frame is slightly blurry where the splices are, but this has nothing to do with stabilizing. (From looking at the vectors it doesn't seem to confuse MVanalyse either.) Each splice is 4 frames long, 2 in each shot, the jerks are longer and earlier/later on.

I'll try Deshaker.

And what is video fred?

If the film jerks(speeds up going though the gate of the projector because of the bad splice) then that is transfered to the video frame. In other words the frame rate of the projector is no longer locked to the video frame rate. Nothing can help except to use a video printer, one frame at a time and a speed~ 1 fps.

VideoFred = http://users.telenet.be/ho-slotcars/s8_index.htm Sorry I couldn't find a English page.

richard

[binba]

That's just what I was trying to say :-) The film doesn't speed up, it moves slightly. In fact, the transfer was done on a scanner so there was no gate. Maybe my word choice wasn't right.

What I would've liked to see is an application-driven interface: instead of specifying the stabilization parameters, I'd want to specify the shake - and let the algorithm figure out what parameters should be used to eliminate that shake.
It could be very useful for my problem, and possibly for other cases of methodical shake. Even though avisynth is probably not the platform to implement such a thing.

[Karyudo]

Hey binba (Drew),

I'm afraid that I can't help with your problem, but I hope you'll forgive a bit of an off-topic side stream of questions:

Where'd you get the telecining done? Home-built rig? Are you in the Lower Mainland (the video downloads from sfu.ca)? Do you do any scanning outside of your own project?

I'll shut up now. If you'd like to keep your thread clear(er), PM me. Thanks!

[rfmmars]

That's just what I was trying to say :-) The film doesn't speed up, it moves slightly. In fact, the transfer was done on a scanner so there was no gate. Maybe my word choice wasn't right.

What I would've liked to see is an application-driven interface: instead of specifying the stabilization parameters, I'd want to specify the shake - and let the algorithm figure out what parameters should be used to eliminate that shake.
It could be very useful for my problem, and possibly for other cases of methodical shake. Even though avisynth is probably not the platform to implement such a thing.

Yes I am interested about the scan too. If your scan inclued the sprocket holes then you do have a reference point to do X-Y-rotation correction

richard

[binba]

Nah I'd gladly talk about the transfer :-) It wasn't exactly a home-built rig, rather a Spirit Datacine 2K (a 2-million dollar machine, from what I was told). Did it in Chicago, the web address doesn't have to do with it.
I might've insisted on them stabilizing the transfer themselves, if not for the huge discount they gave me...

The sprocket holes never get in the frame, it's not THAT bad - about 20 pixel tops, mostly vertical movement. It's all much clearer when you watch the footage...

I tried to use DeShaker but ran into trouble with it, will try posting a new thread for that when I'll have the time.

[videoFred]

Hi Binba,

Quote:
Originally Posted by binba
And what is video fred?
It's a strange Belgian animal, but very gentle!
:D Please do not feed it, it only drinks Belgian Beer:D



Please try this for the first clip:

Avisource("D:\films\Experimenten\AQ-1.avi")
loadplugin("depan.dll")
i=converttoYV12()
mdata=DePanEstimate(i,trust=20)
DePanStabilize(i,data=mdata,cutoff=0.5,dxmax=0)


And this for the second one:

Avisource("D:\films\Experimenten\AQ-2.avi")
loadplugin("depan.dll")
i=converttoYV12()
mdata=DePanEstimate(i,trust=0)
DePanStabilize(i,data=mdata,cutoff=0.5)

It's not perfect, but it improves the clips...
You can play around with the trust, cutoff, dxmax parameters.
Also, at scene changes, you must split the clips, otherwise you get strange effects.

Fred.