I'm not sure how deep this goes, but when I used lwimager to make a 512x512 image, and I look in the FITS headers, I see the center pixels are CRPIX1/2 = 257. Noting also that CRVAL1/2 = 0.
The FITS spec notes that a pixel value references the center of a pixel so that an image starts at pixel value of 0.5 (since we use FORTRAN indexing) and ends at pixel value npix+0.5
So, for a 2x2 image, CRPIX=1.5 and for a 512x512 image it should be 256.5
As an aside, I am currently working on a pipeline to take a model image, a tdl file (fov gives cell size), parset file, etc and produce an lsm file with which I can make an image - the resulting model image should equal the input image. Has something like this been done? Certainly I think it could be useful for debugging this issue, so let me know if the code is useful and I'll share.
I'm not sure how deep this goes, but when I used
lwimagerto make a 512x512 image, and I look in the FITS headers, I see the center pixels are CRPIX1/2 = 257. Noting also that CRVAL1/2 = 0.The FITS spec notes that a pixel value references the center of a pixel so that an image starts at pixel value of 0.5 (since we use FORTRAN indexing) and ends at pixel value npix+0.5
So, for a 2x2 image, CRPIX=1.5 and for a 512x512 image it should be 256.5
As an aside, I am currently working on a pipeline to take a model image, a tdl file (fov gives cell size), parset file, etc and produce an lsm file with which I can make an image - the resulting model image should equal the input image. Has something like this been done? Certainly I think it could be useful for debugging this issue, so let me know if the code is useful and I'll share.