There's actually nothing fancy about this lens calibration. It's
effectively just stitching a single row 360° panorama using more overlap and
more control points than you would normally use for a panorama.
PTOptimizer works on a trial and error basis.. i.e. it involves a certain
amount guesswork. The more information you can supply, the more accurate a
guess becomes, until you eventually provide enough information that will
make a correct answer inevitable. This is the principle on which this
The criteria for collecting images and optimising control points for the
- Correct placement of the lens over its nodal point.
This is critical for a lens calibration as any deviation will result in
different values for a, b and c. You don't need a full on panorama head. A
simple focusing rail will do the job, or even a Metz flash bracket.
- The images have at least 50% overlap
The parameters a, b, c correct for radial distortion in a lens..
i.e. they are used to recalculate the distance of each point from the
centre of the image. To calculate them properly then, we will need
to collect data from the centre of the image to the very edge.
- A large number of control points are used
There's not much point shooting with a lot of overlap if you're not
going to use it. Selecting several control points from all over the
images produces a better overall result. A large number of control points
is also required since a large number of variables will be optimised at
once. Using several points will reduce the risk of a bad result.
- Several vertical lines are defined using the t1 option for defining
This ensures that the parameters will properly correct for lens
distortion. Without these it is still possible to get a good stitch that
may be slightly distorted.