Repository navigation
WCS.all_world2pix ignores a detector-to-image (D2IM) correction present on only one axis #20577
Description
Description
If a WCS has a detector-to-image lookup table on only one axis (det2im1 set, det2im2 left as None), all_pix2world applies the correction but all_world2pix does not. The inverse is off by the full size of the correction, and no NoConvergence is raised. A one-axis D2IM table is a normal FITS distortion setup: it is what you get from a header that has only D2IMDIS1, as in the HST/ACS column-width correction (AXISCORR = 1).
Expected behavior
No response
How to Reproduce
import numpy as np
from astropy.wcs import WCS, DistortionLookupTable
w = WCS(naxis=2)
w.wcs.crpix = [1, 1]
# detector-to-image correction on the x axis only (+0.5 px everywhere)
w.det2im1 = DistortionLookupTable(np.full((5, 5), 0.5, np.float32), (1, 1), (1, 1), (1, 1))
pix = np.array([[2.0, 2.0]])
world = w.all_pix2world(pix, 0)
print("world :", world)
print("round trip err:", w.all_world2pix(world, 0, tolerance=1e-12) - pix)
print("has_distortion:", w.has_distortion)world : [[2.5 2. ]]
round trip err: [[0.5 0. ]]
has_distortion: False
Versions
import astropy
astropy.system_info()# Paste the result here
Activity
mcara commentedon Oct 16, 2026
This is not an issue for HST images as there are no images that have either only D2IMDIS1 or D2IMDIS2. How have you encountered this error?
ChenziqiAdam commentedon Oct 18, 2026
Thanks for the comment, and sorry for the late reply. My statement linking this configuration to HST/ACS was incorrect.
The example sets only
det2im1:all_pix2worldapplies the correction, butall_world2pixskips it becausehas_distortionreturnsFalse. Is a single-axis D2IM configuration intended to be supported? If so, this seems to be a forward/inverse inconsistency. If not, would rejecting that configuration be appropriate?mcara commentedon Oct 18, 2026
That's a theoretical example but I am curious how did you encounter it in real life? What data are you working on?
ChenziqiAdam commentedon Oct 18, 2026
Actually it is not a real example. I was just debugging my code and tried different settings.
pllim commentedon Oct 19, 2026
What we really want to know is how this affects your real work. So what code were you debugging and how was that relevant to this issue?