ginga pick gets central position in 1-indexing

While doing my data analysis for PSF photometry, I found my code always give warning because the ginga pick location (which are fed as initial guess to the true centroid) and the DAOStarFinder centroid differ more than 1 pixel all the time.

To test, I made a perfect Gaussian star at center = (31, 31) to a (51, 51) array.

from astropy.modeling.functional_models import Gaussian2D
g = Gaussian2D(amplitude=1, x_mean=31, y_mean=31, x_stddev=2, y_stddev=2)
test = np.zeros((51, 51))
test_xx, test_yy = np.mgrid[:51,:51]
test += g(test_xx, test_yy)

finder = DAOStarFinder(fwhm=5, threshold=0, exclude_border=True)
test_found = finder(test)
test_pos = (test_found["xcentroid"][0], test_found["ycentroid"][0])
print(test_pos)
# (31.0, 31.0)

plt.imshow(test, origin="lower")
plt.plot(*test_pos, 'rx', ms=10)

test_ccd = CCDData(data=test, unit='adu')
test_ccd.write("test.fits", overwrite=True)

The position by DAOStarFinder is correct: (31.0, 31.0)

But ginga found (32.0, 32.0)

Maybe I guess this is not expected…?

[ejeschke]

Pick reports location in FITS coordinates (1-based indexing) by default, whereas DAOstarfinder probably uses 0-based indexing. Ginga’s internal reporting is 0-based, however, and you can change the value shown in Pick by adding

pixel_coords_offset = 0

to a text file general.cfg in your $HOME/.ginga folder.

[ejeschke]

Might make for some good unit tests, @pllim!