RGB combination

Invocation: nightlight -out rgb.fits rgb red.fits green.fits blue.fits with appropriate filenames for your channels and desired output. You can do arbitrary combinations of channels, e.g. nightlight -out S2HaO3.fits rgb S2.fits Ha.fits O3.fits will put the narrowband S2 data into the red channel, Ha into the green one and O3 into the blue one. You can also pass the same file twice, e.g. to create HaO3O3 bicolor images.

The pipeline for RGB combination has four main phases: 1. preprocessing, 2. reference channel selection, 3. channel combination, 3. color postprocessing

1. Preprocessing

Generally the same as for stacking. Useful steps include:

• Histogram peak scale and location detection. Good for color calibration of the black point
• Star detection. Good for color calibration of the white point
• Alignment. Good to make the color channels line up
• Background extraction. Useful if your subexposures have similar sky gradients. If they differ widely, run background extraction on each subframe

Not so useful steps, which are off by default, include:

• Dark and flat calibration. Typically done before stacking
• Cosmetic correction. Typically done before stacking

2. Reference channel selection

Same criteria as for stacking: many stars, small star size as measured with half-flux radius (HFR).

3. Channel combination

• Each channel is normalized to [0,1]
• Channels are combined

4. Color postprocessing

• Black and white point are adjusted iteratively based on sky background peak in histogram, and average color of detected stars
• Color saturation ...
• Selective color rotation. E.g. for Hubble palette, turning green to gold
• Selective color saturation. E.g. to suppress purple halos.
• Selective chroma noise reduction (SCNR). E.g. to suppress surplus green tints in the Hubble palette.
• Auto curves. Pick where the histogram peak should be, and how wide it should be.
• Gamma. 1.0 does nothing, higher values boost dark areas, lower values darken them. Typical useful values can be 2.0-4.0.
• Post-peak gamma. Like gamma, but acts only on values one or a few standard deviations behind the histogram peak. This avoids scaling background noise.
• Second black and white point adjustment. Just in case manual color adjustments would result in odd star or background colors
• Final black point clipping.