Messier 103 in Cassiopeia

M103, also cataloged as NGC 581, was discovered in 1781 by Pierre Méchain. It is one of the more distant open clusters, 8,000 to 9,500 light-years from our system and measuring about 15 light years across. It holds at least 40 members, the brightest of which appears to be the type M0 red supergiant near the center of the cluster. The brightest star in the area, the triple star Struve 131, is a foreground object rather than a member of the cluster. M103 is about 25 million years old.

Exposure: Total exposure time about 8.6 hours, 154:38:37:30 x 2 minutes LRGB. All bin 1x1. Data collected in September and October of 2020.
Light pollution: SQM ~18.38 (Bortle 7-8, NELM at zenith about 4.5, Red/white zone border.)
Seeing: FWHM of integrated luminance around 2.0 arcsecs
Image scale at capture: 0.6 arcsecs/pixel = f/5.7
Scale of presentation: 1.2 arcsecs/pixel (50% reduction)

Equipment:
Scope: C11 (standard, not Edge) with Celestron 0.63 reducer
Mount: Paramount MX+, connected via ASCOM Telescope Driver 6.1 for TheSkyX, with MKS 5000 driver 6.0.0.0
Camera: SXVR-H694, connected via SX ASCOM driver 6.2.1.17140 (SX 1.2.2 also installed)
Filter wheel: Atik EFW2 with 7x1.25 carousel and Artemis 2.4.3.0 driver
Filters: Astrodon Type IIi LRGB
Rotator: Optec Pyxis 2", connected via Andy Galasso's 0.4 driver (Optec Pyxis Rotator AG)
Focuser: Rigel Systems GCUSB nStep motor with driver version 6.0.7 on stock Celestron focuser
OAG: Orion Thin OAG
Guide cam: Lodestar (first generation). 4 second exposures
Automation SW: Sequence Generator Pro 3.1.0.457
Guide SW: PHD 2.6.7, connected to guide cam via native SXV driver
ASCOM: ASCOM 6.3.0.2831
Platesolving: PlateSolve 2, failover to local Astrometry.net 0.19 server
Collimation: Metaguide 3, using ASI120MM connected via ZWO Direct Show driver 3.0.0.2
Processing Software: Pixinisight, Affinity Photo, Photoshop CS2

Processing Workflow by Workspace in PixInsight 1.8.8:

1. Calibration
Calibration with WeightedBatchPreProcessing with flats and bias, using Cosmetic Correction with a master dark
Blink to preview and reject a few frames
Weighting and registration with WBPP

2. Stack and Mure Denoise
Image Integration on each channel
Mure Denoise on each channel
RGB Combination for RGB frames
Dynamic Crop
Dynamic Background Extraction

3. Luminance Linear Processing
(I skipped PSF-based deconvolution – see the next step)

4. Luminance Stretching
STF Transform, copied to Histo Trans and slightly adjusted. This was easier than my standard approach of combining HT and Curves Trans.
Parametric Deconvolution. I’m finding that on clusters this gives better results than a PSF-based Deconvolution on the pre-stretched luminance. Stars have sharp edges, while deconvolution on the pre-stretched image often results in star halos.
TGV Denoise

5. RGB Linear Processing
Photometric Color Calibration, using Average Spiral Galaxy white reference

6. RGB Stretching
Histo Trans
Boost blue saturation with Color Saturation
Histo Trans
Curves Trans
Boost blue saturation with Color Saturation, using a clipped luminance mask to apply only to the stars and not the background

7. Color Combination
LRGB Combination of Luminance and RGB images

8. Background Subtraction
 a. Create an image of the background:
   1. StarNet++ to create an image without most of the outlying stars. This leaves a bright blotch for the core of the cluster.
   2. Modify the starless image in Photoshop:
     a. Use the Healing Brush and CloneStamp tools to remove halos, leaving only the background
    b. Apply a heavy Noise Reduction filter so that noise is not removed during the subtraction process.
 b. Subtract the background image from the original image (using Image>Apply Image) to remove optical artifacts (rings in the image) and any remaining messy clumps in the background. Use a mask to prevent the bright core of the cluster itself from being subtracted, and adjust the offset to get the right background brightness.
 c. Save as TIFF and move back into PI

9. Star Reduction
I used a modified version of Adam Block’s star reduction technique:
 StarNet to create a new “Starless Image”
 Extract two copies of luminance from the Galaxy Image, then apply a 6-layer MLT, unchecking the residual layer, to one to create a rough star mask.
 Binarize to eliminate everything fainter than star cores
 MorphTrans using dilation to enlarge the stars
 Edit the mask with CloneStamp to exclude any background galaxies
 Convolution to blur star edges
 Pixel Math: subtract luminance image from blurred star mask so that cores are excluded from mask, and on ly halos are represented in the mask = “Halo Mask”
 Apply Halo Mask to main image, then run PixelMath to use Starless Image where halos otherwise would be

10. Final
Curves Trans to slightly boost color saturation
ICC Profile Transform to sRGB
Resample at 50% scale
Save as JPG