< Previous | Contents | Manuals Home | Boris FX | Next >
Determining the 3-D position of a curve is at the mercy of underlying mathematics, just as is the 3-D camera analysis performed by the rest of SynthEyes. Because every point along a curve/flex is equivalent, there is necessarily less information in the curve data than in a collection of trackers.
As a result, first, flex analysis can only be performed after a successful normal 3- D solve that has determined camera path and field of view. The curve data can not help obtain that solve; it does not replace and is not equivalent to the data of several trackers.
Additionally, the camera motion must be richer and more complex than for a collection of trackers. Consider a flex consisting of a horizontal line, perhaps a clothesline or the top of a fence. If the camera moves left to right so that its path is parallel to the flex, no 3-D information (depth) can be produced for the flex. If the camera moves vertically, then the depth information can be obtained. The situation is reversed for a vertical line: a vertical camera motion will not produce any depth information.
Generally, both the shape of the flex and camera path will be more complex, and you will need to ensure that the camera path is sufficiently complex to produce adequate depth information for all of the flex. If the flex is circular, and the camera motion horizontal, then the top and bottom of the circle will not have well-defined depth. The flex will prefer a flat configuration, which is often, but not necessarily, correct.
Note that a simple diagonal motion will not solve this problem: it will not explore the depth in the portion of the circle that is parallel to the motion path. The camera path must itself curve to more completely identify the depth all the way around the circle — hence the comment that the camera motion must itself be more complex than for point tracking.
Similarly, tripod (nodal pan) shots are not suitable for use with the curve & flex solving system. As with point tracking, tripod shots do not produce any depth information.
Flexes and curves are not closed like the letter O — they are open like the letter U or C. Also, they do not contain corners, like a V. Nor do they contain tangency handles, since the curvature is controlled by SynthEyes.
Generally, the curve will be set up to track a fairly visible edge in the image. Very marginal edges can still be used and solved to produce a flex, if you are willing to do the tracking by hand.
©2024 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.