• astropy
• ccdproc
• pyregion
• photutils
• image-registration
• joblib

All scripts defined below use argparse for processing command line arguments. Each program can be run with the -h flag to print a description and example usage (e.g. python stack.py -h).

Each script also takes the optional --c flag to clobber/overwrite files on write.

Finally, each script also has an optional -njobs flag that allows for multi-process execution. The parameter to this flag is the number of jobs (processes) to run in parallel. If -1, all CPUs are used. If 1 is given, no parallel computing code is used at all, which is useful for debugging (this is also the default mode for all scripts). For njobs below -1, (n_cpus + 1 + n_jobs) are used. Thus for -njobs -2, all CPUs but one are used. For more information on this, see the joblib documentation.

python stack.py NGC4565/raw/*.fit -imgtype bias -o NGC4565/reduced/MasterBias.fits

Alternatively, stack.py can accept one input filename (and all the other command line flags, too). This one filename is then assumed to contain a list of input files, one per line.

python stack.py NGC4565/raw/*.fit -imgtype dark -exptime 60 -o NGC4565/reduced/MasterDark_60.fits
python stack.py NGC4565/raw/*.fit -imgtype dark -exptime 120 -o NGC4565/reduced/MasterDark_120.fits

Optionally, add the --n flag to normalize by EXPTIME. This will be performed regardless in `proc.py'.

python stack.py NGC4565/raw/flat*.fit -o NGC4565/reduced/MasterFlat.fits --normw --wolly -maskfile masks/wolly_mask.reg

Optionally, add -njobs -1 to parallelize.

The array has many bad pixels. Use LACosmics to clean.

python clean.py NGC4565/raw/NGC4565_R-0* -odir NGC4565/reduced/ -sclip 3

This can be quite slow; add the -njobs to parallelize if possible on your machine.

python proc.py NGC4565/reduced/NGC4565_R-0*.fit -bias NGC4565/reduced/MasterBias.fits -dark NGC4565/reduced/MasterDark_120.fits -flat NGC4565/reduced/MasterFlat_*.fits -odir NGC4565/proc/ -maskfile masks/bad_pix.fits

Specifying -maskfile attempts to mask bad pixels during the CCD reduction process. However, this does not always produce clean images during the dark subtraction step. Adding the --fixpix flag will interpolate over the bad pixels, yielding nicer looking images, but this also adds significant computation time. We recommend using the -njobs option to process multiple images in parallel. Avoid using every CPU, since the re-gridding will fill up each CPU.

python split.py NGC4565/proc/*.fit -odir NGC4565/split -maskfile masks/wolly_mask.reg

Using the wolly_mask.reg ds9 region file, split the images into ordinary (A) and extraordinary (B) beams.

python imalign.py NGC4565/split/NGC4565_R-0* -odir NGC4565/align --c -m extended -xy 512 204 -box 100 100

We can align all the images using either a specified point source (-m point -xy INT INT) or through 2D cross-correlation (-m extended). It may be useful to make cutouts of the images (using the -xy and -box flags) rather than attempt to align over the whole array. The alignments can easily be parallelized.

• Alignment procedure is not perfect, particularly on noisy data.
• Q-U pair subtraction is not complete.
• Organize scripts into pipeline mode (maybe using luigi?) rather than running each step individual.