DIGITAL
3D CINEMATOGRAPHY
Conventional 3D: A programme (for example - Maya, 3D Studio MAX etc.) which creates wireframe objects and renders them, adds lights, shadows, textures, animation etc. This is still flat or two dimensional to the viewer when displayed on a monitor. (See the page "3-D Animation")
Stereoscopic 3D: A simultaneous left and right eye view of an object or scene presented to the viewer which creates three dimensional image in the brain. Objects protrude from the screen and scenes appear to have real depth and distance.
The fundamental basic for stereoscopic 3D-images is the use of two cameras to capture left and right eye images, positioned to mimic the human eye's stereo vision through seeing two subtly different angles of the same scene.
To match the human stereo vision obviously requires two cameras with their optical centres spaced approx. 2.5 inches apart - just like the human eyes. However, much like the way that color and contrast in a film image (digital or celuloid) is exxagerated to enchance the theatrical experience, this interocular distance, or more correctly - interaxial separation, is often increased to exaggerate the stereo effect, and add impact to the viewed image.
There are two different approaches to setting convergence. The first is to "toe-in" , or converge, the cameras such that both cameras point at the object to be placed on the screen plane (zero parallax). Although the issue is that each camera will suffer different keystone convergence, or vertical parallax distortion, making the images hard for a viewer to "realise" visually if these keystoning effects are not fixed during postproduction.
The alternative is to shoot parallel and set the convergence in post. However, changing the convergence point in post is making unnatural change as the relative parallax isn't also changed. That's why, I constructed my stereoscopic camera rig with the option of toe-in the cameras.