| Home | Gallery | Other | Contact |
In this area of the website: 3D | About Me | xxxxx | Exhibitions |
 

How 3D works

 

This is the text of an article that I wrote many years ago.

3-D images

Does anyone know how to create 3-D images in Infini-D or other programs? I'm talking about the kind of images that need those cool glasses...

To create 3D (stereoscopic) images you need:

Two pictures of the scene taken form slightly different positions.

A transmission process that ensures that each eye gets the image that was intended for it.

That's the short version.

However that doesn't give you any clues as to how to achieve this. For the purpose of this brief discussion, we'll assume that the image is to be generated and shown via a computer VDU - although I'll note some other more effective transmission processes.

Two pictures of the scene taken from slightly different positions.

There are three main things to think about when photographing/rendering an image for 3D reproduction.

They are focal length, interaxial distance (The distance between the two cameras) and convergence. All of these can be varied to suit the image, many people preferring to exaggerate them to increase the apparent depth.

Focal length.

If you were using a 35mm camera, the 'classical' focal length on a 24 x 36mm frame (Landscape) is around 85mm. This gives a perspective that closely approximates that of the human eye. The typical 50mm lens is slightly wide, but having said that, most lenses between 40-100 will not cause undue perspective distortion.

Interaxial Distance

Classically 2-3 inches, but many 3D photographers use 4-5 inches to increase the apparent depth. For special uses this can be increased or decreased - but you should be aware that doing this makes the image look smaller or bigger - for example, there isn't any human depth perception of space or planets.

Convergence

Convergence is the way your eyes point at the object that you want to look at. Thus you go very slightly boss-eyed. When you're making stereograms, the point of convergence is usually where the picture plane (screen in the case of a VDU) is located. To make a stereo nice to look at, the most important thing is best located here. Making it stick out is a sure way to give the viewer a head ache. There should be no independent up and down motion of the cameras. I use the 'point to' command to establish a correct convergence. Using zero convergence can make the eyes diverge (go outwards) which is very unnatural. However, zero converged stereo-pairs are quite common and can be partially corrected in the transmission process by overlapping the images. A good test for any stereoscopic image is a sphere, try modelling and producing something that feels the right depth and size as a tennis ball.

A transmission process that ensures that each eye gets the image that was intended for it.

Basically, if you want to look at your image on a computer screen there are two ways: Anaglyphs (Red 'n' green to you and me) and using either a stereoscope pushed up against the screen or free-sight.

Anaglyphs

Once you have produced your two images they will need superimposing. This is easiest done by using photocopy. Simply reduce them to greyscale and place one in the red channel of an RGB image (Left is traditional) and then place the other in the green channel. The blue can be left.

Wot no colour? There are so-called full-colour anaglyphs, using (usually) red and cyan. you can obviously try these - but frankly they don't work very well as they rely on the eye to combine the colours, and you can't get a full spectrum anyway from red and cyan. Besides which, there is a difference in brightness between the transmission of the red and cyan filters - all of which adds to eye-strain. After a time the colour will fliker (Known as colour bombardment) A monochrome picture with Red and Green is fairly comfortable to view. When you have the images in PS, you can adjust the horizontal convergence slightly. Make sure that the most important parts of the picture are the most exactly superimposed. If you've used zero convergence this a good way to make the picture more comfortable to view - most old time 3D pictures do something similar to this as a matter of course. (It's called setting the stereo window.) If the picture doesn't seem right, try turning the glasses around.

Other viewing methods:

I have successfully used a stereoscope pushed against the screen of my PB180 (downsampling to 256 didn't do the image much good). Putting the two images side-by-side and going boss-eyed worked well. - or you could output to transparency and project with polarised light - but that's another topic!

 

Pages in this section:

3D Film

How 3D works - a loose explanation

3D Cameras

3D Images:
Pearl

 

 

 

 

 

Last update September 30, 2005