Jamie Smith

Jamie Smith

Woodworker. Explaineer. Gadget Geek.

RGB & SHO image processing workflows, 2024 edition

8 minute read

In the past few years the tools used to process astro-images have made a generational leap. Tools built into PixInsight, like WBPP (Weighted Batch Preprocessor), have simplified calibration and integration, while other free tools, like Generalized Hyperbolic Stretch (GHS), and SetiAstro’s helpers are making processing in PI easier (once you learn them). Other, paid tools, like RC-Astro’s *XTerminator tools have amplified the leap. In this post I’ll walk through a high-level overview of how I process images. It’s repeatable, pretty streamlined, and I think that you can get pretty good results without a lot of theory or math.

Workflow Overview
Workflow Overview

I’ll start out by reiterating that this is my 2024 workflow, because odds are that it will change. More new tools, new techniques, and more learning.

If you want to cut to the chase, you can (always) download my process icons by cloning my pixinsight-icons GitHub repo or downloading them as a zip file

High-high-level overview of the process

For my primary imaging rig (I should document that somewhere, it’s gone through a few revisions), I tend to shoot both narrowband and broadband images, with per-filter exposure times as follows:

  • Lum (180s)
  • Red (300s)
  • Green (300s)
  • Blue (300s)
  • Sii (600s)
  • Ha (600s)
  • Oiii (600s)

If I’m just processing a single night of data I might not go through the entire process, but even then it’s still similar to the flow I use for final images.

A preliminary step is to use WBPP to calibrate and cosmetically correct frames (if that’s not already done), and to align, register and integrate them. I turn off the autocrop and embedding of the rejection maps in WBPP– My raw frames are already 120mb, I don’t need them bigger :grinning:

Once I’ve got the calibrated and integrated master frames, I proceed to the flow. Note that for things like globular clusters I might not do the SHO path, unless there’s something interesting.

  • Open, tile, rename, and stretch the applicable L/R/G/B and S/H/O frames
  • Process the RGB frames together
  • Process the SHO frames separately
  • Merge the RGB stars with the SHO starless
  • Final tweaking

Breaking down the flow overview

Baseline setup, with SHO & RGB frames from the Heart Nebula
Baseline setup, with SHO & RGB frames from the Heart Nebula

For the first “Open, tile, rename, and stretch” step I use the scripts that I created to get to a baseline view of what I’ve got. In this case you can see what that looks like for the Heart Nebula that I recently processed. Once I’m to this point I can start the flows.

Disclaimer on any of the menus/script locations: Some of the tools aren’t built in, but if you include the repos in the appendix they should be in the right places.

The RGB Workflow

image-right

For each of the workflows I’ve included a screenshot of the PixInsight process icons that I use for the flow. I pretty much start at the top and work my way down. For scripts that don’t provide the ability to create a new instance (the little triangle), I put a “No Op” (NOP in the set) as a placeholder, with a descriptive name and, in some cases, details in the icon description.)

  • DynamicCrop (note that I crop the SHO and RGB frames with the same instance of dynamic crop)
  • Combine (L)RGB channels (I just open this and run it)
  • Rename the combined image as RGB (I still use an ImageIdentifier icon for this, since it’s static)
  • Once I’ve got the RGB image I usually iconize the individual channels and drag them off to the side
  • at this point it’s probably going to be way off on color, so I do an unlinked STF stretch
  • Run the script AutoLinearFit on the combined RGB, usually set to use the red frame (alternatively, manual linear fit, before the combine)
    • A note here: I’ve had issues with BlurXTerminator if I use the Lowest Mean setting onAutoLinearFit
  • Another unlinked STF stretch here doesn’t hurt
  • Gradient Removal: I run through a few different methods to see which works best for a particular image. I go through these tools to compare:
    • Seti Astro’s “Automatic DBE”
    • GraXpert
    • The new GradientCorrection tool
    • I generally clone the RGB image during the gradient experimentation, but when I’ve chosen, I apply it to the actual RGB so the image history stays intact
  • I drop on the BlurX_CorrectOnly icon, which is BlurXTerminator, with the “correct only” option selected
    • This is supposed to help with the plate solve (which usually gets borked in the crop)
    • I actually don’t always do this step unless the next ones fail
  • I drop on the PlateSolveForSCC icon to get the solution. Even though the crop loses data, this usually still works
  • Then comes the actual SPCC, or Spectophotometric Color Calibration, for which I use one of two flavors, depending on the object type:
    • For spiral galaxies, I drop on the SPCC_RGB_ASG icon (the “ASG” is for “average spiral galaxy”)
    • For everything else I use the SPCC_RGB_PF (photon flux) variant
    • An important note for this step is that you MUST UPDATE the process to set the correct sensor and filter types for your equipment. Once you do that you can swap out the icon for your info
  • CHECKPOINT: We should now have a gradient-free, color-calibrated image
  • Perform a full BlurXTerminator correction, by dropping on the BlurX_Full icon
  • Drop on the RGB_AddStarless_2_ID, to rename the image to RGB_starless
  • If needed I drop on the FixMagentaStars
  • I’d then try noise reduction with NoiseXTerminator by dropping on the NoiseX_Tweak icon
    • I call it “tweak” because, with some images, I’ve got to jiggle the handles a bit to get a good look
    • Like above, I make sure that the final execution is on the same image with the history
  • We’re now getting in the home stretch! I extract the stars with StarXTerminator by dropping on StarX (which has the “Unscreen” option checked)
  • Finally, I carefully make the STF-autostretched stars stretch permanent, by dragging the STF state onto the HistogramTransformation process, then applying it to the stars

At this point I might be done with the RGB side, unless there’s something cool in the starless version. If I did it all right, I’m ending up with two new images: an RGB_starless and RGB_starless_stars. I might play with the starless version and rescreen the stars (definitely will if there aren’t any narrowband frames), but I usually move on to the SHO side.

SHO Workflow

Right off the bat: there are some icons in this that I do not use for final images, but I keep them around for quick sanity checks of data. A good example are the SPCC steps in the SHO path

image-right

Like the RGB flow, I’m starting out with open narrowband frames, but I don’t merge them right away, but rather do a little mental loop… For EACH of the SHO frames, I do the following:

  • Perform some magic by dropping the SHO_BlurX_Full icon on them
  • Experiment with the various gradient removal techniques, similar to the RGB flow (but, instead, on the discrete frames)
  • OPTIONALLY, Stretch the frames with GHS (this can be done later, after they’re combined)

Then I combine the images into an SHO image, mapping Sii → Red, Ha → Green, and Oiii → Blue. **Don’t Panic. It’s gonna be green. Very **green… There are a couple different options, but the straight channel **mapping has been working for me as it goes through the workflow

  • With my newly combined image, I drop on the Rename_SHO icon to change the image identifier to be SHO
  • I run the script Script → Toolbox → AutoLinearFit on the new SHO with the green channel as the reference
  • I perform NarrowbandNormalization on the SHO image, selecting the SHO palette, turning on the live preview, and tweaking the channel controls and adjustments (if you don’t know what they do, just crank them to 0 or 11 to see the impact)
  • then an STF autostretch
  • I skip the SPCC options
  • In preparation for StarX, I drop on the SHO_AddStarless_2_ID, to rename my SHO to SHO_starless
  • I drop the NoiseX_TWEAK_SHO icon onto the image, then decide if I need to tweak the settings (I find that, once you’re familiar with your images, you can get pretty consistent results)
  • then, nuke the stars!
    • This should leave you with SHO_starless and SHO_starless_stars images
    • if you have an RGB_starless_stars image, the SHO_starless image is probably no longer needed
  • If you haven’t stretched yet (AKA, the image is still linear), run a GHS on the SHO_starless image
    • Experiment on where you do this step (before vs after combining). It’s easy to overstretch if you do it earlier in the process, but you can get a lot more detail that way
  • Depending on the subject, I try the “DarkStructureEnhance” script, which is in the Script → Utilities menu

The rest is usually experimenting with the overall look. Maybe enhancing the hue or saturation with curves or a “local histogram equalization”.

The last step is screening back in the stars, but I usually safe off a copy of the SHO_starless image first. If the RGB_starless_stars are too big I try to shrink them a bit.

You can use basic pixelmath for that, or the new ImageBlend script (I’ve got icons for both of those). If you’ve been following the same conventions I have they should just work. If you HAVEN’T, change the icons to match your flow.

This Heart Nebula was the first image that I took through this flow start to finish, and I was very pleased with the results (and I even got a top-pick nomination for it on AstroBin)

Heart Nebula SHO with RGB stars
Heart Nebula SHO with RGB stars
Heart Nebula

Useful?

If you want to provide feedback, hit a snag, or just want to let me know that you’ve used the icons (or my scripts), file an issue on the GitHub repo, or let me know on AstroBin or instagram.

Appendix

Repositories

The repos that I use in these flows:

https://www.ghsastro.co.uk/updates/
https://www.cosmicphotons.com/pi-modules/narrowbandnormalization/
https://www.ideviceapps.de/PixInsight/Utilities/
https://www.cosmicphotons.com/pi-scripts/starreduction/
https://www.cosmicphotons.com/pi-scripts/imageblend/
https://raw.githubusercontent.com/setiastro/pixinsight-updates/main/
https://raw.githubusercontent.com/jamiesmith/pixinsight-repo/main/

And the others that I have configured

https://www.cosmicphotons.com/pi-scripts/nbcolourmapper/
https://elveteek.ch/pixinsight-updates/ez-processing-suite/
https://foraxxpaletteutility.com/FPU/
https://www.cosmicphotons.com/pi-scripts/bfke/
https://www.skypixels.at/HVB_Repository/