< Previous | Contents | Manuals Home | Boris FX | Next >
Lens Settings
Lens Master Calibration
Once the input data for calibration has been set up trackers or linesyou're ready to set up the calibration processing itself.
NOTE : Be sure to save your current scene file before starting calibration. Not only is that useful in case something goes wrong, but the file name is used as a default lens name, so make it informative about the lens.
Start the calibration process with Shot/Lens Master Calibration menu item, or simply by starting the Lens Master Calibration script within the Lens section of the scripts.
This will cause the initial lens calibration dialog to appear, which is where you tell SynthEyes what properties of the lens you want it to compute as part of the calibration. The input controls are laid out pretty much in the order you need to think about them. When values aren't being computed, the supplied value is used; when the value is being computed, the supplied value serves as a starting point for finding the correct value.
Note that in addition to the descriptions below, you can reference the tooltips of the various fields while running the script.
Lens Name Override. Text. You can specify a different lens name here, it is used for the preset name and image map names, so just use simple safe characters such as a-z, 0-9, underscore, and dash.
Type of Calibration. Drop-down. Here you specify what kind of calibration you are doing, based on the type of imagery and how you've set it up, ie Fixed Spacing (grid/checkerboard), Straighten Lines, Random Dot, Known XYZs.
Pans a full 360deg. Checkbox. This applies only for Random Dot calibration; you should always shoot full 360+ shots, so you should always have this on.
Type of Lens Calculation. Drop-down. Here you can enter what kind of lens calibration you want done: Linear, Inverse Linear, Anamorphic, Fisheye, Equisolid, Orthographic, Stereographic.
The Linear type is the usual SynthEyes type for small to moderate amounts of distortion, were the distorted position is computed from the undistorted position. Inverse linear is the same type of calculation, except that the undistorted position is computed from the distorted position. The inverse linear type is used by packages such as Nuke that aren't concerned with undistortion, just with undistorting an image as rapidly as possible.
The Anamorphic selection is for use with anamorphic lenses, of course. When this type is selected, an additional control panel will appear for it; this is discussed later.
NOTE: Lens Master Calibration does not currently support or use the same lens models as the new “Advanced” lens models introduced in SynthEyes 2304.
The fisheye, equisolid, orthographic, and stereographic selections are all types of fisheye lenses; you can read about the details on the wikipedia article on fisheye lenses. If you know it, you can select it directly. Otherwise, you can try each of the selections and see what gives the best fit. If you do that, however, do not permit any additional distortion terms (see next control) while comparing them. First determine the type without distortion, then you can do the calculation with distortion.
Distortion Equation. Drop-down. Here you can select the type of distortion coefficients computed, from none, quadratic, cubic (and quadratic), quartic with no cubic (includes quadratic), or quartic (with cubic and quadratic). Note that Nuke appears to use the inverse linear option with quartic and no cubic coefficient. (The values would have to be adjusted to match Nuke's.)
Permitting the higher-order terms such as quartic to be computed allows more complex distortions to be modeled more accurately, but if there isn't as much distortion, it causes spurious, unnecessarily complex, coefficients to be computed that are reacting to noise, rather than actual distortion. We recommend starting with the simplest quadratic distortion, and comparing that to the more complex versions. When spurious coefficients are present, it's typical to see positive and negative pairs in cubic and quartic coefficents, ie one canceling the other.
You should stick to simpler distortions with fisheye lenses, especially if you are computing the field of view, as permitting a complex distortion allows the shape of the curve to be changed from the nominal shape, which is what allows the field of view to be computed at all.
Compute lens center . Checkbox. When checked, the lens center position (U/V values from -1 to +1) will be computed. This is only possible when there is substantial distortion present. Otherwise, see the section on vignette-based lens center measurement. When the box isn't checked, the lens center values in the Lens center U and V fields will be used (which default to zero).
Compute FOV. Checkbox. When checked, the lens field of view will be computed. This is only possible for fisheye lenses, or using the Random Dot or Known XYZs calibration methods. It's only possible for fisheye lenses (and not the usual linear lenses) because the
specific mathematics of the lens serve as a reference, if you like. (Random Dots and Known XYZs do make the calculation possible.)
Estimated Field of View. Number. Whether a FOV is computed or not, you should always enter an estimated field of view into this field, as it is used as a starting point for further refinement. You can use the Field of View Estimator to compute it for you for linear lenses, based on the nominal lens and camera parameters. Fisheye lenses require some care, as they are typically specified for a full image circle, which often extends past the image rectangle, so you can't really tell how much of the full-circle image is being used.
Field of View Estimator. Button. Brings up a small panel to help you estimate the field of view from lens focal length and sensor size.
Lens Radius. Number. With the default value of 1, the lens radius says that the field of view corresponds to a radius from the image center to the center of the left/right edges of the image, ie U = +/-1. When processing full-circle fisheye lens images, the non-black optical image won't extend to the edge of the digital pixel array. The Lens Radius must be set to the radius of that used image circle, the boundary of which corresponds. See also the following Diagonal FOV setting.
Compute sensor tilt. Checkbox. When turned on, this enables some additional parameters to be computed that are intended to compensate for tilt of the camera sensor chip versus the optical line of sight. If computed, it can only be rolled into a lens UV image map. In general this should not be necessary; it is here mainly for experimentation.
Correct for rolling shutter. Checkbox. This is useful only for the Random Dot measurement; it allows you to enter a Rolling Shutter value and have it used to compensate the tracker positions for the camera motion. Even though the camera motion is typically slow, due to the precision of the Random Dot calculation, we do highly recommend supplying an accurate value and correcting for it for random dot calibration.
Rolling Shutter. Number. A -1 to +1 number determined elsewhere, used here only for Random Dot calibrations.
©2023 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.