Package for denoising NIRCam calibrated images

Description:

1/f noise introduces random vertical and horizontal stripes inside the JWST NIRCam images.

denoise_NIRCam has been developed to remove stripes from each image row/column subtracting its median value, obtained masking objects and DQ-flagged pixels . The object masks are obtained running SExtractor (Bertin & Arnouts 1996) and dilating the resulting segmentation image, applying a careful procedure to dilate segmentation depending on ISOAREA of objects:

• a 3x3 convolution kernel and the desired dilation d in pixels is applied for objects with ISOAREA lower than a threshold
• a 9x9 convolution kernel and a dilation of (m x d) pixels is applied for objects with ISOAREA greater than the threshold, being m a desired multiplicative factor to apply to the desired dilation d.

The denoising is performed over the entire columns and can be executed separately for each amplificator row or over the entire rows. The entire rows modality is recommended in presence of very extended objects.

Processing parameters are passed as command line options and the defaults are expressed inside brackets. Boolean options do not need arguments, calling or not calling a boolean option means 'True' or 'False' respectively. The defaults of boolean options are expressed in curly braces.

Here is the help message with the complete list of the options:

denoise_NIRCam.py [ -h | --help ]

denoise_NIRCam.py [ -V | --version ]

denoise_NIRCam.py [-o OUTSUFFIX] [-a] [-d DILATION] [-f FWHM] [-i ISOAREA] [-k] [-m MFACTOR] [-p] [-s DETECT_SIGMA] [-v] image1 image2 ... imageN [@filelist]

-h, --help            show this help message and exit
-o OUTSUFFIX, --outsuffix=OUTSUFFIX
                      The suffix for output images [.dn.fits]
-a, --amp             Correct background stripes for each amplifier? If
                      False, the corrective medians of background stripes are 
                      computed along the entire horizontal axis {False}
-d DILATION, --dilation=DILATION
                      The dilation in pixels to flag sources [5]
-f FWHM, --fwhm=FWHM  The fwhm in pixels [4]
-i ISOAREA, --isoarea=ISOAREA
                      The minimum area to mask extended objects [4000]
-k, --keepall         Keep all flags in DQ layer? If False,
                      only pixels that are marked as 'DO NOT USE' are
                      discarded [False]
-m MFACTOR, --mfactor=MFACTOR
                      The factor used to multiply the dilation of extended
                      object mask [4]
-p, --purge           Purge temporary intermediate files? {False}
-s DETECT_SIGMA, --detect_sigma=DETECT_SIGMA
                      The sigma used to detect sources [4]
-v, --verbose         Verbose output [False]
-V, --version         Print program version number, then quit.

The use of the '-k' option is highly recommended, in order to discard outliers in the median estimates as much as possible. The sigma for detection should not be chosen too low, to avoid spurious detections (e.g. parts of stripes or even entire stripes, which have to be removed and not masked) nor too high to miss many detections of real sources.

A typical Linux command line, using a dilation=5 pixels with mfactor=4x and a minimum isoarea=5000 pixels to mask extended objects, looks like:

denoise_NIRCam.py -i 5000 -k -d 5 -m 4 -p -v jw_cal.fits 

Here is a typical NIRCam image before (left) and after (right) the 1/f denoising:

If you use the code for publishing purposes, add the following papers among citations:

• Paris, D., Merlin, E., Fontana A. et al. 2023: arXiv e-prints, arXiv:2301.02179. [https://arxiv.org/abs/2301.02179]

• Merlin, E., Bonchi, A., Paris, D. et al. 2022, A&A, 938,827 [http://doi.org/10.3847/2041-8213/ac8f93]

• Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 [https://ui.adsabs.harvard.edu/abs/1996A%26AS..117..393B/abstract]