Camera Aray HELP

[Levan Bakhia]

Please help me here. so, here is a situation,

I have purchased 60 canon 350Ds and sync them perfectly, I have set up a system so that when I capture a frame, all pictures from the cameras are transfered to computer via a special software that will name frames according to camera position and so on. (well as you would guess, I am trying to achieve effect from matrix) So, my system works absolutely fine by now, in terms of functionality, but the problem that I have now is that cameras are not taking the same pictures, I mean that if I take one and same lens and use it with different cameras, and I mount the cameras to the same place, which is steady and not moving, so when I change the cameras, leaving the lens the centre of the picture is shifting. I am not sure if you understand what I mean, but it is impossible to align cameras correctly if the cameras are not taking the same kind of frame. and the result is shaking. So, do you know any of the ways to correct that? I do not know if it is sensor which is not always the same location in the camera, or is it a body of the camera and the lens fits in defferent place but the fact is that it is not aligned.

Any help, suggestion is so much welcome, please, I have invested over 80K in this and more than a year and I am almost at a perfect system in terms of a functionality and how it captures and transfers files, but now I have this problem which I can not solve.

[Giroud Francois]

this problem is the same for everybody.
you need to to treat the pictures as a shaky video and stabilize it.
an easy way to do it is to use marks (bright leds, fluo marking) that you can use as motion track.
the only solution is software.

[Levan Bakhia]

Well, that is for the 2d correction of the shift, but the problem is that because when you aim all 60 camera centers to the same spot in space, because cameras have shifts inside, you end up not centering the cameras actually. it is too complicated to explain in words, but bottom line is that you end up in shifts in perspective, and how can you stabilize picture perspective with software?


Ohh, and what about "everybody", are there many who do that?

[Giroud Francois]

yes everybody who tried the "bullet time" effect was confronted to that problem.
I have seen not so bad sequences produced by teenager.
Most of people solved by framing large, then align by centering et zooming when necessary.
a second step usually require to interpolate additional pictures to get smooth slow motion.
I doubt you can mechanically build a system that is definitely set to give properly centered picture (even with laser).
if between two pictures you still find that something doesn't match, it is probably that some camera was too late, so the logical sequence in time is not correct.
It not uncommon to have to picture taken not to create the sequence, but to provide a texture to CG sequence, that recreate the volumes. it is easier then to play in 3d as you want.
Definitely, bullet-time effect is not an point and shoot process and require a lot of post production work.
from pure logical, you can not get different perspective if a camera is using same lense, same position, same zoom factor.
At least not from having only few pixel offset.

[Levan Bakhia]

Hmm, not calming at all. So, even when they filmed Matrix or Swordfish, they had the same problem? Can you recommend software for interpolating pictures in between?

Again, how do you correct 3d shift of perspective?

[Ben Winter]

because cameras have shifts inside

This is the one part I don't understand. I helped The Big Freeze do a commercial shoot for Discover Card (see my website) and learned a lot about their method, and I could possibly help you if I only knew what you meant by this.

If the perspective is off then the camera is mounted in the wrong place, not framed incorrectly.

[Giroud Francois]

motion perfect is a nice soft to interpolate new frames from the existing ones.
for the defect yo have, the best is to provide 3 or 5 pictures in sequence showing the problem.

[Levan Bakhia]

Ben, Giroud

Thanks for being helpful, I will prepare a page by tomorrow, where I will explain everything, so that you can go an see yourself, frams, videos, explanaittions and results. So, please check back tomorrow, I will have it posted, if I achieve the good result, I will put up a web page where I describe the whole process for other to share.

So thanks you, and give me time to post a question more clearly.

[Levan Bakhia]

OK, here is my question on this link http://www.filmingeorgia.com/question


Please note that there are 5 pages, on top are the links, don't pay attention to the page layout of course I this is may first web page :*)

[Giroud Francois]

Sorry cannot open your quicktime.
The processe to caliber pictures should be simple.
first you have to set zoom factor and level equally on all camera.
to do this you shoot a large horizontal ruler and make sure it looks the same size on all camera. By the same time, you set focus.
Do not trust the picture you see in the visor or on LCD, because it is not the picture that will hit the sensor, but the one that hit the reflex mirror. if the position in the sensor has small variation (and it will) this will give you wrong result. The only way is to have a laptop to check the size in pixel of your ruler in the hires picture.

Then when all camera are ok, you need to to make sure that the subject is centered more or less the same on all camera (again a centermark on the ground glass would help and a mark on the subject laser pointer ?)), make sure to keep level.
Since the sensor is not mounted at exactly the same place in all camera, you cannot avoid a little shift beween picture or even variation in zoom factor, but this will minimize the loss in border's when you will crop pictures.

shoot your scene.
load all picture in your image editing software. take a rectangular selection, centered on the mark you made in the scene (or a small detail visible on all pictures).
this way you can crop a centered picture of same size in all the original shots. If nicely done, you should have a good result.
there is a utility in a software called "reality studio- photo vista 3D object" that allow to center pictures with the arrow cursor. http://www.iseephotovista.com/galler...3dobjects.html
you should try it. I think quicktime VR on Apple has the same tool.
Ideally you could write a software than can remember the correction for each set of camera/lens, so you can take a row of picture, feed the system with and it would correct automatically. This can be done in photoshop, but you need to be an ace with batch and scripting.
All this is probably more or less the same you have already done, but if you fail getting proper result, you should have missed something.

[Cole McDonald]

Stabilizing the footage can also be done easily, either in shake, combustion or after effects (even FCP has a stabilizer built into it you could use-not as good though). On the matrix, they also hand generated morphs between each frame to make them blend seamlessly. There is free software out there for doing this.

[Ben Winter]

First:

Originally Posted by Levan Bakhia

In a bigfreeze video that I found on Discovery Card on Ben Winter’s page (thank you Ben) I see that guys have this tirangle markers layed around on the shot, so Ben can you tell me in brief what it is and how it works?

These are nothing more than focus markers. We used 28mm lenses on all so the DOF was fairly deep, but we still needed solid, straight objects to focus on to make sure the right things were in focus in the frame.

Second:
I'm not fully understanding your issue. I think I have an idea of what the issue is but I'll get to that later.

Please clarify for me:
You say that you're changing lenses and the view is different each time you switch out a lens/camera. That's to be expected as the build tolerances for lens/camera mounts are not exact in the x/y direction and will have some change from lens to lens or camera to camera.

But why are you switching out lenses/cameras? Can't you just keep the lenses and cameras alone and focus/frame them individually? Please tell me you have each camera on it's own adjustable ball head...

Third:
Are you framing the cameras correctly? You keep speaking of a shift in perspective but all your diagrams are showing are a simple shift in framing instead, which is correctable in post-production.


Also, your videos have not been encoded correctly. They contain no information. They should be many megabytes in size.

because cameras have shifts inside

Like I asked before, I need to know what you mean by this. It does not make sense.

Ultimately, I really just need to see a video of your latest attempt. And two or three consecutive jpeg images from adjacent cameras. That would help immensely.

EDIT: WAIT! Are you using zoom lenses? DON'T! Each lens is not manufactured to be at the same focal length when 'zoomed out' or 'zoomed in'. You can only use prime lenses with these setups! The zoom ring settings (and EXIF data) for zooms is not that precise and it is not uncommon for zooms to be shorter than the stated focal length. Your footage may reflect that of many cameras at different focal lengths.

[Levan Bakhia]

----But why are you switching out lenses/cameras? Can't you just keep the lenses and cameras alone and focus/frame them individually? Please tell me you have each camera on it's own adjustable ball head...----

Yes, all cameras are on its own adjustable ball head, and I am switching lenses/cameras just for experiment, to see if there is a difference in camera mounts, sensor or whatever.

----Are you framing the cameras correctly? You keep speaking of a shift in perspective but all your diagrams are showing are a simple shift in framing instead, which is correctable in post-production.----

Well, yes, you can correct the shift easily in post production but what happens is that when you are framing the camera, aiming to the same spot in space, you really are making incorrect alignement because, on the first hand it is not accurate.


----Also, your videos have not been encoded correctly. They contain no information. They should be many megabytes in size.----

I will correct that, sorry I think it is not about encoding it correctly, it is rather about not publishing it correctly.



--- Like I asked before, I need to know what you mean by this. It does not make sense. ---

Within next two hours I will film a video, of myself, explaining what I mean, and I will upload it so that you can see it. because just by typing text it just makes no sense, I have to show what I mean.


and yes I use zoom lense, which I also thought not to be the best option.

[Giroud Francois]

zoom lenses are ok , it just makes the adjustement longer.
if you make sure that an object has the same size on every camera (measuring a test object on each camera) there should be no problem.
the previous advice i made about an horizontal scale to do that was wrong.
with perspective such a scale will look longer or shorter depending you are just in front of it (with the middel camera) or on a border.
if you scale is vertical, this should be better (while not perfect).

[Ben Winter]

Originally Posted by Giroud Francois

if you scale is vertical, this should be better (while not perfect).

This requires that all cameras be the same distance from the subject.