< Previous | Contents | Manuals Home | Boris FX | Next >
Determining Field of View
In order to create a match between your model and the imagery, not only does the mesh have to be correct and properly positioned, but the correct camera field of view must be determined and configured (the "perspective" must be right). There are a number of ways to do that.
NOTE: Though we talk about the "field of view" throughout SynthEyes, it is technically an "angle of view" measured in degrees, rather than a linear measurement.
The most direct way to determine the field of view is to measure it directly, typically with a tape measure. For example, point the camera at a wall. Looking through the camera (viewfinder or monitor), measure horizontally across the entire field of view right at the edge of visibility, to obtain the width. (Do not use a monitor that has "overscan" for this measurement!). Next, measure the distance to the wall. Finally, compute the field of view, using the Camera Field of View Calculator, found on the Lens submenu of the main Script menu. But of course to do this, you must have recorded the measurements during the shoot.
You can use the pinning tool to compute the field of view in post-production, by turning on the "FOV changes" button on the pinning toolbar, and with enough (accurately placed) pins, the field of view can be determined with moderate accuracy. (Accuracy will be greater when the object covers a larger portion of the image.)
If the scene contains a number of parallel and perpendicular lines, you may be able to use the single-frame alignment toolkit; see "Tripod and Lock-off Shot Alignment" in the SynthEyes user manual. This method is particularly relevant when you want to determine the position of a stationary camera relative to items in the scene (a desk, say), before tracking the moving object(s).
If the camera is moving, or panning and tilting on a fixed tripod, you can use a regular trackers and a camera solve to determine the camera path and an accurate field of view over the entire shot. This method produces the most accuracy, since it is based on the entire shot, but is likewise more complex (producing more data, ie the camera path, as well). This method is the only choice for handling shots where the lens zooms during the shot.
As a very last resort, you can attempt to use the focal length value reported from the camera lens. Not only is this number only approximate (manufacturers cite tolerances of +/-several percent), but it is only half a number! It is useless without a precise sensor (back plate) width. Unfortunately, the sensor width value is difficult to obtain exactly. Even when manufacturers report them, they are approximate (two digits of precision), and you don't know how much of that sensor area corresponds to the pixels produced. A total error of 5% corresponds to almost 100 pixels of an HD image, which would be a huge mismatch. Fortunately the errors are relative, not absolute, but even still focal length values must be viewed as a last resort.
If you will be measuring something during the shoot, measure the actual linear field of view of the camera, from edge to edge, at a specific measured distance from the camera. Then use Script/Lens/Field of View Calculator.
©2024 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.