Fix the incorrect method of setting the distortion model flag

[demul]

The previous way of setting distortion model is wrong.
For example, if you want to use thin-prism model, but you only set cv2.CALIB_THIN_PRISM_MODEL as True, not cv2.CALIB_RATIONAL_MODEL and cv2.CALIB_TILTED_MODEL, the calibration result sucks.
Thin-prism model must have other polynomial coefficients which is enabled by setting cv2.CALIB_RATIONAL_MODEL and cv2.CALIB_TILTED_MODEL as True.

I fixed above problem.

[oliver-batchelor]

• False is an odd expression

I presume this would suffice?

  rational=cv2.CALIB_RATIONAL_MODEL ,
  tilted=cv2.CALIB_RATIONAL_MODEL | cv2.CALIB_TILTED_MODEL,
  thin_prism=cv2.CALIB_RATIONAL_MODEL | cv2.CALIB_TILTED_MODEL | cv2.CALIB_THIN_PRISM_MODEL,

I have never used rational models myself.

[demul]

Right, Your way is better for masking bits.

[demul]

I commit to satisfy your change request,
And it seems that, "The previous way of setting distortion model is wrong." is different by OpenCV version.
In some late versions, just setting cv2.CALIB_THIN_PRISM_MODEL flag only is enough to enable all 12 coefficients.
But in some old versions, setting cv2.CALIB_THIN_PRISM_MODEL flag only is not enough. It just enable 4 coefficients(s1, s2, s3, s4) not other 3 coefficients(k4, k5, k6)

And there was a mistake! In, OpenCV, The tilted model(14 coefficients) is considered as exetended version of thin_prism model(12 coefficients).
So

model = struct(
      standard=0,
      rational=cv2.CALIB_RATIONAL_MODEL,
      thin_prism=cv2.CALIB_RATIONAL_MODEL | cv2.CALIB_THIN_PRISM_MODEL,
      tilted=cv2.CALIB_RATIONAL_MODEL | cv2.CALIB_THIN_PRISM_MODEL | cv2.CALIB_TILTED_MODEL,
  )

is right code.