Camera/Lens Distortion Calculation?

[starscream]

Hello,

I do not have a strong astronomy background, but I do work with machine vision cameras and have to calibrate intrinsic parameters (distortion, focal length estimation, etc.) for cameras + lenses occasionally. However, I'm confused by some of the terminology. For instance, I believe 'calibration' in astrometry is closer to meaning an extrinsic calibration (localizing our camera relative to some other reference images). I'm wondering if there is a straightforward method for computing lens distortion parameters using the output information provided by astrometry.net? For instance, if we know a constellations position in the sky and the relative position between the stars within that constellation (from some well-calibrated and undistorted reference image), then we should be able to determine the distortion error of my camera for most of the pixels containing that constellation (to within some tolerance). A good distortion calibration would likely benefit from multiple images, and images which contain constellations in the outer perimeter of the image, but for now I would like to know if I can get the data from my astrometry output to determine distortion error. Thank you in advance for your help, and I can try to provide more details if you need more information! 

[Dustin Lang]

Hi,

You're right, what we call "calibration" is basically the parameters of a pinhole camera ("TAN"gent-plane projection in WCS-speak), ie, the RA,Dec position on the sky of a reference point, the pixel coordinates of that reference point, and the 2x2 matrix that describes dRA,dDec by dx,dy.

Lens distortion is called "distortion correction" or "SIP" (for the name of one of the pseudo-standard representations), or "tweak" (astrometry.net's name for the algorithm to fit it)  Currently we only do single-image fitting for it, but yes, it would be possible to use many images to improve the fit.

The SIP results are stored in the wcs.fits file, which is a FITS header (ie, text file with archaic formatting); the SIP standard is defined here, https://fits.gsfc.nasa.gov/registry/sip.html

SIP is a general polynomial (no specialization for the kinds of distortion patterns you might expect from optical systems).

cheers,

dustin