Jump to content

Translating data from a technical camera to a cg-camera


nisus
 Share

Recommended Posts

Hi All,

 

What is the best way to translate information from a technical camera to a cg-camera?

 

Basicly we have some great photos taken by an architectural photographer with a technical camera, but standard camera-match doesn't work at all...

What modifiers can be used to adjust the cg-camera so that it perfectly matches the enhanced perspective of the technical camera?

 

rgds

 

nisus

Link to comment
Share on other sites

Basicly we have some great photos taken by an architectural photographer with a technical camera, but standard camera-match doesn't work at all...

 

What you probably mean is that the photos are perspective-corrected, either shot with a PC lens on a small format camera, or a 4x5 box camera with a bellows lens. Those aren't used much, so probably the first. Either way, it's not that hard to figure out.

 

You must have your CG camera with a matched Zheight for both camera and target. Pan the image (not tilting the camera) to match the film shot. In Lightscape and Cinema this is extremely easy, in Max/Viz I guess you use the camera modifier trick.

 

Beyond that, its trial and error to find the amount of lens offset you need. That assumes you have already matched the location and height of the shot--but you always must do that.

Link to comment
Share on other sites

You must have your CG camera with a matched Zheight for both camera and target. Pan the image (not tilting the camera) to match the film shot. In Lightscape and Cinema this is extremely easy,...

 

That is exactly what I do in Max/Viz - carried over from my Lightscape days. :)

Link to comment
Share on other sites

Hi Joske,

 

The solution is not really in the camera modifier because we need to be able to place the camera first...

 

 

Hi Ernest,

 

I do that technique regularly, so there is no big issue about that... BUT... with this project, it's not only LEFT and RIGHT that has to be correct, also the intermediate facade has to be perfectly correct. Basicly we have to change the interior in a very long facade and add a few things on the exterior of the facade too... In my opinion, the photo looks stretched, but not in a lineair way...

I'll try to post an example asap.

 

rgds

 

nisus

Link to comment
Share on other sites

In my opinion, the photo looks stretched, but not in a lineair way...

nisus

 

then the only solution would be to start of with a 2D vector facade plan.

convert it into .eps data to import in photopaint (or photoshop)

chop up the photograph and start puzzeling away on top of the vector lines in photopaint. (giving each piece a different stretch as needed)

this would be in my exp the only way to get rid of the non-linear stretch

 

(there is a 'perspective correction' tool in photopaint that takes the non-linear stretch into account up to a certain level, but it will only just allow minor changes)

Link to comment
Share on other sites

then the only solution would be to start of with a 2D vector facade plan...

 

Let's not get too complicated. Remember that if you have the same camera location in XY and the Z height right, and the camera target centered correctly, the views should match.

 

Assuming you have the view matched pretty well (as above) then you can render it with a little wider view angle than your photo, and use the render as a base to modify the photo!

Link to comment
Share on other sites

Hi all,

 

For those interested and who would like to give it a shot, here is a part of the file, with two different maps. One for the camera-correction-modifier-approach, one for the levelled-Z-camera ,-)

 

http://www.ams.be/download/CGA/AMS_CameraCorrection.zip

 

The goal is to cameramatch as correct as possible so that every cg-window exactly fits it photographic counterpart.

For the visualisation, we have to keep as much of the original photograph as possible (in the facade), replace the windows and interior and add a few attributes to the exterior.

 

(Maybe I should make an unchallenge-topic out of this???)

 

rgds & tnx

 

nisus

Link to comment
Share on other sites

Nisus,

 

You need to model with some depth greater than just the facade, like the White tower in the background and put camera points to that. Some "z" axis depth to the xy facade would help a great deal for cam match in max.

 

If you have an image taken from the opposite angle and or different views, preferably with the same camera / lens /focal length, I could give Image Modeler a try, generate a reference-facade & camera for max.

 

 

WDA

Link to comment
Share on other sites

Hi WDA,

 

Well the original model has more depth, but the main goal to show for me was to show (through eyeballing) that a huge problem exists because the photo looks/is stretched. If you give it a quick try, you'll see it immediately...

 

Anyway, I'll provide the photo from the other side and the 3ds soon.

 

rgds

 

nisus

Link to comment
Share on other sites

Hi WDA,

 

Well the original model has more depth, but the main goal to show for me was to show (through eyeballing) that a huge problem exists because the photo looks/is stretched. If you give it a quick try, you'll see it immediately...

 

Anyway, I'll provide the photo from the other side and the 3ds soon.

 

rgds

 

nisus

Ya, I tried it before posting, for about 45 minutes- Manually that is. It's a tough nut to crack!

 

I'll look at the file/s.

 

rgrds

WDA

Link to comment
Share on other sites

The photo's perspective is distorted. Look how the vertical lines are. They're not paralel, they shoot up, and with different rates. This one is a spherical perspective, not a conical one. I can't draw a straight line in the upper side of the buildings. The more to the left you get, the more the line shoots upwards from where it should be if the perspective was conical.

 

It'd be almost impossible to use Camera Match to match it.

 

On the other hand, it's a 2D perspective only (façade view) so it's very easy to manually match the camera.

 

Alexander

Link to comment
Share on other sites

wda,

 

tnx for effort anyway ,-) have you found what is wrong with the photograph (imho this is the cause of the prob)

 

rgds,

 

nisus

Yes it's the photo most certianly.

 

1. There is a perception of the twisting of the facades' planes, or a very crooked construction.

 

2. It looks like a crop of radial distortion (lens).

 

Image modeler did'nt like it much, two of the far left upper markers, became unusable almost like thier normal flipped there but un selectable, really wierd stuff. The model was about 1/2 kilometer long and 20m tall when exported to max. (pushed the geometry to fit the lens lens) If you can get a human face to model in in the program...., really started to think it's the photo.

 

The tall image with the white matte gets very close to getting the horizontal to align. Going on that same thought maybe by adding more matte to the righthand side and using for the background image- the focal point (actual Camera) & Max's (center of the render) may get closer. I think that is where the problem lies. The photo was cropped and the lens distortions' center is not the center of the photo. The perspective is always off, if you...could just... move something...... a little more .....to get the perspective to close quicker..... Like adding about 125 pixels to the righthand side of the bitmap?

 

Look at the the 04_01_18.jpg and the flat04_04_16018.jpg comapre them, the lens distortion and perspective would be identical identical if the the flat...jpg was not cropped, I'm guessing.

 

That's enough for tonight... the shelves here look like they have developed camera distortion.

 

WDA

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...