in 3D Equalizer
Oct 09, 2012
November 2014 3DE distortion video:
Matrix, in F3, with the plate or a photographed distortion grid
Manually adjusting the distortion in F3, with the overlay grid, with the plate or a photographed distortion grid
Parameter Adjustment window, which uses the 2D tracks as the basis for distortion
F5, adjusting the distortion and focal length value at the bottom of the view port, while watching the "Hitchcock zoom" alignment of the LIDAR.
Of course with the new
Lineup window, you should even need to check anything in Maya,
fine alignment should be visible in 3DE F5. Notice that for a
dense LIDAR, 3DE has the very useful hidden line removal, not
generally available in Maya, unless you use the Mental Ray
Contour function for software render.
Also, the Anamorphic Squeeze parameter has nothing to do with a Cinemascope/Panavision lens with a non square pixel aspect ratio of two. So this value on a spherical or anamorphic lens will often be around 1.
FYI, here are the distortion parameters available in the big four tracking softwares. 3DE has the largest number of adjustments, although with great power comes great responsibility, i.e. the extra parameters are very useful for difficult shots, but can be confusing.
3DE: degree 2, degree 4, plus three or four trims
3DE Anamorphic has 22 parameters, only for emergencies
pfTrack: degree 2, degree 4, plus anamorphic squeeze (very different than anamorphic)
SynthEyes: degree 2
Boujou: degree 2, LCO
All have LCO
adjustments as well.
Theoretically the Brownian/Newton values in 3DE and pfTrack should be the same, but they do not give identical values. So pfTrack needs the pfBarrel node in Nuke and 3DE needs the WETA/3DE undistort node in Nuke. There is also a SynthEyes undistort node available for Nuke.
"Modern" grid will determine focal length, but it typically
has a high deviation error.
So the "Classic" straight grid is often better, according to Rolf.
In the old days before the Matrix, the grid would need to be positioned exactly parallel to the camera, which is almost impossible to do in practice, outside a laboratory.
But with a "straight" Matrix, theoretically the grid can be keystoned slightly away from parallel and still give great results, much better and easier than the old fashioned traditional grid, without a Matrix. But again, only an angled grid will determine focal length.
But for focal length, there are other issues. An angled Matrix shot will require the grid to be placed very close to the camera and the focus setting will be wrong. A plate is usually shot at about ten feet away, but because the printed grid is small, the grid shot usually has to be focused at a distance of about 2 feet. Thus, the calculated FOV is off slightly, because of lens breathing. So I like to stop the camera down to t/16, focus at the "correct" 10 foot distance and let depth of field keep the image sharp. And with an angled grid, there are often checkerboard squares that aren't even covered at the edge of frame, or are too small or out of focus to Matrix accurately.
The other methods of determining focal length still work. If you have accurate LIDAR in F5 Lineup, you can "Hitchcock zoom" the focal length and low order distortion. Or you can do a Fine Adjust of focal length, etc., in the Parameter Adjustment window, as always.