Hyades Open Cluster

Hyades Open Cluster in Taurus

Full Resolution

The Hyades Star Cluster has never really fascinated me much until I got into the hobby of deep sky imaging. As a bright set of stars in the night sky it’s commonly known as The Bull from ancient mythology. I didn’t know just how much star forming activity was going on in this region of space until my attention was directed to the Large Dark Molecular complex in taurus, and even then I hadn’t a clue how far it extended across the sky. This picture depicts some of the further reaching arms of that dark series of nebula. Inside it most prominently is Aldebaran, the bright yellow star. To the upper left is NGC1647 a far younger open cluster. On the upper edge in the middle is Hind’s Variable Nebula(NGC1555) and Struve’s Lost Nebula (NGC1554.) Probably my most favorite region of this image is the Sh2-239 reflection nebula sitting in extremely dense molecular cloud LDN1551.
* Sh2-239 in LDN1551

Follow Me on Instagram | Astrobin

Equipment:

  • Celestron CGEM Mount

    • Self tuned / hacks to get guiding stable include:

    • Intentional offset polar alignment so dec always pulses in one direction

    • Balance “west” heavy (rather than the recommended east) so that the ota “falls” onto the gear teeth rather than get “lifted”

    • Factor Reset hand-controller daily (to prevent cgem from being possessed and forgetting where the meridian is on subsequent night)

    • Dither in RA only

  • Rokinon 135mm f2 at f2

  • asi071mc PRO at -15 C

  • Astronomiks L3 UV/IR filter

  • Widefield Rig + AF3 by Deep Sky Dad

Acquisition:

  • 136x120s at unity gain (4 hours) taken on 2020-02-14 and 2020-02-21

Session sequenced using Nighttime Imaging in Astronomy: N.I.N.A

All pre-processing and post-processing was done in PixInsight and final touches added with Affinity photo. Full details below.

The resulting image is a combination of the following steps:

  • Inspected all subs for bad images with Blink, discarding subs containing clouds

  • Calibrated all subs with a master flat and master dark

  • Used subframe selector to weight all images based on the following weighting

1
2
3
4
5
6
7
8
9
10
11
12

(10*(1-(FWHM-FWHMMin)/(FWHMMax-FWHMMin))

+ 20*(1-(Eccentricity-EccentricityMin)/(EccentricityMax-EccentricityMin))

+ 30*(1-(Median-MedianMin)/(MedianMax-MedianMin))

+ 10*(SNRWeight-SNRWeightMin)/(SNRWeightMax-SNRWeightMin)

+ 10*(Stars-StarsMin)/(StarsMax-StarsMin))

+ 30
  • Selected the best sub from subframe and blink to use as a reference frame for aligning and integration

  • Cropped the stacking edges of the integrated image

    • Color Calibration
  • Dynamic Background Extraction

    • Color Calibration
  • Background Neutralization using 5 small preview windows aggregated as a background reference

  • Color Calibration |> Solved |> Photometric Color Calibration

    • Color Calibration
  • RGBWorking Space to 1,1,1

  • Noise Reduction was done using TGV Denoise with a low contrast mask and an autostretched local support and MMT with a very protective luminance mask

    • TGV Denoise and MMT Denoise
  • Stretched using Arcsinh Stretch followed by Masked Stretch

    • Arcsinh Stretch
    • Masked Stretch

Luminance Processing

  • A synthetic luminance was extracted prior to the stetching of the RGB data for separating luminance processing L

  • Stretched using Masked Stretch and Histogram Transformation

    • Masked Stretch with Histogram
  • Created a Starless version of the luminance using starnet++ and a StarsOnly image by subtracting the starless from the luminance Starless-L

    • Starless Version after applying StarNet++
  • Performed two rounds of Local Histogram Equalization on the starless image

    • Kernel Radius 128 | Contrast 4 | Amount 0.100 | 8-bit | Circular

    • Kernel Radius 256 | Contrast 8 | Amount 0.030 | 8-bit | Circular

  • Performed large scale sharpening and noise reduction with Multiscale Linear Transformation 6 levels

    • Larger scale sharpening with Local Histogram Equalization and Multiscale Linear Transformation
  • Recombined the enhanced Starless with the Stars only image with Pixel Math Starless+StarsOnly

    • Re-add the stars back to the starless

Bringing the Enhanced Details back into the Color Data and Final Steps

  • Used ChannelCombination in CIE L*A*B mode to apply Luminance to Color

    • Re-apply the luminance details back to the color image
  • Stars in this image were reduced by using two rounds of morphological selection with a luminance based contours mask created with the following steps:

    • Extract a new luminance from the color image

    • Apply the following pixel math expression to a new extracted luminance to create a contours version using the StarsOnly image created earlier: StarsOnly/$T-$T

    • Morphological Selection with Contours Mask to reduce star intensity

  • Curves transformation was used to increase contrast and saturation with a luminance mask and shift the tones towards the golden brown color

  • Over saturated stars were magenta and were fixed by inverting and applying SCNR green before re-inverting

    • Morphological Selection with Contours Mask to reduce star intensity
  • Final round of curves to balance the blues

    • Morphological Selection with Contours Mask to reduce star intensity
  • An ICC Profile was applied to enable Black Point Compensation



Constructive criticism is welcome. Let me know what you think! How can I improve?