Cineversity

Hello everyone,

I feel I stuck at this point and I really need your help
I’m trying to create C4D camera to match a photograph. Here’s the input:

1. Landscape photograph shot on Canon 5D Mark III.
2. The camera sensor is 36x24 and I guess film gate in C4D is 36mm

&

1. Portrait photograph shot with the same camera.
2. Should I use the film gate of 24mm?

Thank you very much!

Alexander.

Hi Alexander,

The Gate is supposed the horizontal measurement of the format (in C4D, in other apps it can be different). So your question is spot on. It should be 24mm in portrait mode, considering a “Full-Frame” sensor (Canon’s web site specifications: Image Format Approx. 36 mm x 24mm (35mm Full-frame)).

I assume you like to match the image as it would be a background image. (For Camera Mapping that is a different case! No change needed here, just make certain you get not a cropped image—press the calculate button then in the Attribute Manager >Texture Tag.) If the render camera has to fit to the image you like to use, set the Gate to 24mm.
But that is not the whole story. To get in this format completely, you need to set up the “Render-format” accordingly: which is 2:3 or in this case the aspect ration of 0.6667 instead of 1.5 (3:2) which is used for landscape mode with that format.

You can do you test quickly by rotating the landscape camera (36mm) around 90º (R.B)—this should result in the same image as if you use the portrait aspect ratio and 24mm.
(See attached scene.)

I have marked the camera as well the render-settings accordingly, so switch both and you should get the same image, of course with the landscape 90º, which is needed to rotate in post. Note that this little test is designed for the camera object relation, the default light will not change at all. Again, the idea is here to give you a method to proof the way, not a suggestion to work at any time with a 90º camera. I have used here in both cases a 50mm lens. Please note that the sensor is not exact 24x36mm so far my knowledge, and lenses are not always exact, especially zoom lenses differ from their values. e.g.,  the 24-70mm/F2.8—v1 vs v2 have a different field of view!

All the best

Sassi

Hi Dr. Sassi!

Thank you so much for the prompt reply. You were right I’m matching camera to the BG picture, for I’d like to create a a different room for the object (which is green screen keyed). I acme to the same conclusion about 24mm, but really wanted to have your professional opinion. One thing I was kinda confused with while trying this 24mm gate theory is “35 eqv. of focal lentgh”. The moment I set the gate to 24mm it changes the “35 eqv of FL” and I tried to match it 35mm eqv (say my EXIF says 55mm I was trying to set Focal length to 36.5mm) - which obviously was a mistake which proved by your example of two cameras. No matter what the gate the FL has to be the same and to disregard “35 eqv of FL” correct?

Sorry butI didnt get the “the 24-70mm/F2.8—v1 vs v2 have a different field of view!” part. What are v1 and v2?

Thank you very much again! Great help!

Alexander.

Hi Aleaxander,

You’re welcome.

I hope I get your question correctly. The Gate needs to be set up first, as that is not an option to be changed in a DSLR for example. The Lens length is not connected to the sensor size and stays always the same. In fact most lens can support even a larger sensor. The resulting projection of a lens is typically round (Some Cine-lenses have a cropping, top and bottom, to avoid flare). This round circle is always a little bit greater than you need it. If not you see some problems. Also here, exceptions are given, like with an 8mm fisheye, where the circle is smaller than 24mm to get the full circle. On the edges you can see a lot of problems. The largest circle is typically given with Tilt Shift lenses, so you can move the lens aways from the sensor and keep the sensor covered anyway. I digress, but I like to give exceptions as well instead of incomplete generalization. Another exceptions is given with lenses created specifically for smaller sensors, same focal length, but the provided coverage (circle) is smaller.

Given the fact that the lens focal length is not changing, the gate or sensor size takes not always full advantage of a lens. As it takes then only a part of the circle it looks like a “longer” lens, hence people name this crop factor. Long story and ask four people you might get even more explanations. I use one lens on my RED, and the resolution that I have set up creates a smaller use of that same lens. Lens and sensor distance aren’t changing. Based on the image size settings, a 200mm shows up like in a 2K image as 650mm “lens, compared to an Full-frame image”. But in 5K as a 260mm, always the same lens.

After all these examples, the focal length of a lens is not equal to the filed of view that the sensor finally provides, the smaller the sensor, the less it uses the lens. Same with the gate. The lens set the maximum angle of view, the gate doesn’t change the options of the lens, but it uses only a little bit of it.

If you take a video with your camera this is not 36x24! It might be rather 16x9 perhaps, which would change the whole set up.

The 24-70 example was given as I imagined you have investigated the best lens options for your camera. The newer version (v2) seems to give a wider range than the label promises. So to assume to just take one end of the lens and have 24 or 70 mm as a stable value might limit your precision.

My advice to match thinks, as far as possible: take always any measurements that you can get on set. The camera high, the view angle, etc. If in a green screen studio, measure the distance of tracking markers, or C-Stands.  Measure the distance of anything that is in the picture and how far it is to the camera. The more you know the better you are of. Take a lens grid to match the lens distortion.

OK, I have perhaps provided too much information. ;o)

Good luck with you project

Sassi

Thank you for the detailed answer - it’s indeed full of useful info. Unfortunately - the first and the most important rule of the VFX shoot - “grab all data/measurements” didnt happened that time and I have bunch of pictures with little to no valuable information - sort of relative e.g. a person height and pieces of tile on the floor.
It’s kinda looks good if to look at one picture but if to take a look at the series of composited pictures the “built-in” objects sort of not consistent… I mean looks not real size wise or other parameters wise.
My question is - have you ever tried to match Canon 5D Mark III with Canon 24-105/4L by any chance and what do you think about EXIF info of this set of camera+lens e.g. focal length, subject distance? It is reliable?
I’m probably going to rent one and do a couple tests to figure out the game.

Thank you!

Alexander.

Hi Alexander,

I have discussed this a little bit here:

http://www.cineversity.com/vidplaytut/2d3d_background_and_camera_mapping

If you can’t use the camera calibrator for that (I don’t know your image), then you need to have an un-cropped image for that.

This might give some clues. Re-verse engineering is the key the that. Even with nothing in the scene, the middle point is always the target point. This target point is important.

I never have used this lens, but I have used over several decades quite some lenses. Each is different of course. What you need to know is how much distortion is given in it, with the exact setting used. Yes the EXIF data helps.
Especially in “GreenScreen” cases, where an identification is harder that in a standard room for example.

If you know the hight of the camera, and roughly if the camera was looking up or down, that might help already. To be honest, its not really helping to not see the image.

To find the camera position is not simple but I have done it a lot of times. If you have other images from the set (while nothing has changed of course, you might get some support in Syntheyes for example. In the old days Image modeler was helpful for that as well.

All the best

Sassi