< Previous | Contents | Manuals Home | Boris FX | Next >
What we would call a lens these days—whether it is a zoom lens or prime lens, or fisheye lens—typically consists of 7-11 individual optical lens elements. Each of those elements has been precisely manufactured and aligned, and by the nature of this, they are all very round in order to work properly. Then they are stacked up in a tube (along with gears and other mechanisms), which is again very round, to form the kind of lens we buy in the store.
The important part of this explanation is that a lens is very round and symmetric and has a single well-defined center right down the middle. You can picture a laser beam right down the exact center of each individual lens of the overall lens, shooting in the front and out the back towards the sensor chip or film.
With an ideal camera, the center beam of the lens falls exactly in the middle of the sensor chip or film. When that happens, parallel lines converge at infinity at the
exact center of the image, and as objects get farther away, they gravitate towards the center of the image.
While that seems obvious, in fact it is rarely true. If you center something at the exact center of the image, then zoom in, you’ll find that the center goes sliding off to a different location!
This is a result of lens de-centering. In a video camera, de-centering results when the sensor chip is slightly off-center. That can be a result of the manufacturer’s design, but also because the sensor chip can be assembled in slightly different positions within its mounting holes and sockets. In a film camera, the centering (and image plate size) are determined solely by the film scanner! So the details of the scanning process are important (and should be kept repeatable).
De-centering errors creates systematic errors in the match-move when left uncorrected. The errors will result in geometric distortion, or sliding. Unless rendering packages render images with a matching de-centering, mismatches are guaranteed.
The lens center can be determined as part of the solve, by a camera calibration, or via other physical methods (see the Camera Calibration manual for details on the latter). While having the solver determine the center is conceptually simplest, the center can only be accurately determined if the shot contains sufficient motion to reveal it. If the shot doesn’t contain enough information, that is generally equivalent to saying the exact center doesn’t matter, and the exact image center can be used. In shorter shots with many straight lines where an accurate center may be necessary, those straight lines can be used for a manual calibration. In other cases, perhaps the physical methods might be indicated.
Note that after the Lens Workflow has been run, the lens center is always in the exact center of the image: the images are pre-padded to make sure that is the case.
The centered center minimizes makes the downstream exports simpler and more reliable.
©2024 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.