The brain is used to being able to “fuse” the two separate views of the world seen from each eye into a 3D perceived scene. However we have a closed-loop system and the eye muscles constantly pull convergence and focus depending on the region of interest. When presented with two separate image streams originating from the left and right camera systems of a 3D rig, the eye – brain coupling has to deal with potentially unusual errors, inconsistencies and paradoxes to try and fuse the images into one. If any one or more of the many possible camera misalignments is out, then for some people who view the 3D, their brain and eyes will struggle to pull-in the images and make sense of them – hence the problem.
Relative to each other, there are 5 axis camera alignments that must be set very accurately and held during shooting. This is why 3D rigs are (or should be) so solidly built, to ensure camera alignment can be set-up and rigidly maintained. There are the 3 relative rotations: pitch, roll and yaw. In 3D the yaw (Y axis) rotation is particularly important as it sets the convergence of the optical axes. Any vertical misalignment must be eliminated and needless to say, both cameras should be on the same focal plane.
The other axis is the interocular, the distance between the lenses and is more an artistic decision than a technically fixed point. There can be wide range of variation chosen, depending on the scene and the intended viewing conditions. Added to 5 critical camera axes, there are the variations and inconstancies between the lenses. Further more, there are many additional factors inherent in each camera to affect matching such as white balance, gamma and image stabilisation – which should be off.
In some opinions there is an attitude of assuming everything can be ‘fixed in post’. This may not always be possible such as with lens flare or lost genlock. It is certainly not optimum, not just because of the time and cost but because any fix to a stereo pair’s geometry must involve having to throw away part of the source images or compromise the dynamic range and resolution. For these reasons initial camera set-up, alignment and monitoring before and during shooting is critically more important in 3D than 2D. Its not just one extra dimension to make good 3D – it’s a dozen or more.