StarTools

Here is an image of NGC 6744, a nice spiral galaxy in the Southern Hemisphere. It compares well with the APOD published at this URL: https://apod.nasa.gov/apod/ap140808.html

In my on-going attempt to upgrade my skills with StarTools, I'm not sure I'm succeeding very well with stars. The transitions to the cores seem very abrupt to me. And I'm having a hard time leaving Curves behind. It's such a useful tool in making final tweaks to the tone curve.

Your advice would be welcome.

Russ Carpenter

Hi Russ,

Could you be a little bit more specific about what you would like advice on?

The transitions to the cores seem very abrupt to me.

Do you mean in terms of color or stellar profile?

If you mean stellar profiles, then please understand that stellar profiles are not something to "hide" with an over-exposing stretch. Stars should not bloat; stars are not little discs or blobs. They are point lights diffracted by your optics and the atmosphere. As such a good image with a good stretch should show this diffraction pattern perfectly. The image should not further over-expose cores of stars that do not over-expose in the source dataset, and should not show the cores of any stars that do over-expose in the source dataset "bleed" into neighbouring pixels. E.g. a good stretch only does this;
E.g. it perfectly brightens parts that were not visible before. It does not change the appearance or "size" of stars or their cores.

I'm having a hard time leaving Curves behind. It's such a useful tool in making final tweaks to the tone curve.

Bezier curve manipulation is a very archaic and sub-optimal way of establishing a baseline global stretch. Leaving these tools behind (particularly as a means to "tweak" your image) is crucial if you want to make the most of your signal.

I'm saying "establishing a baseline global stretch", because such a stretch should only serve as a starting point for further local dynamic range manipulation. Doing further global stretching-base tweaks after local dynamic range optimisation, invalidates all your local dynamic range optimisations and all the assumptions you (and the algorithms) used to make those tweaks. Make no mistake about it; these old tools are the gateway to "tweak-upon-tweak" hell (both from a usability and signal processing/fidelity point of view) and are the most important things to unlearn asap.

ST's eninge goes through great lengths to offer you the ability to make those refinements at the most opportune time (for example, allowing you to do Decon after stretching). Everything hinges - in terms of the decisions you make based on what you see visually, and what the algorithms measure in terms of SNR - on your baseline.

Think of image processing in StarTools as sculpting; a global stretch is like establishing the rough shape of the marble, local dynamic range manipulation is the act of defining finer and finer features (Contrast for large-to-medium local dynamic range, then HDR and Sharp for the smaller stuff). One comes before the other. Few sculptors tack on a heap of clay across their sculpture again as a final "tweak"...

I hope that makes sense!

Hi Ivo,

Your comments make perfect sense. Your support of the StarTools community is remarkable and is one of the main reasons I want to return to StarTools.

The graphics and text relating to star profiles nailed it. I can proceed with more confidence.

As to Curves, I, like most other astro imagers, have spent decades with Photoshop and Photoshop equivalents (in my case, it's Affinity Photo). And, of courses, Curves are also a fundamental tool in PixInsight. Thus, there are some deep habits to "unlearn." I'm now used to the idea of abandoning CurvesTransformation, HistogramTransformation, and ArcsinhStretch for front-end stretching, but I'm not yet very good at making small downstream corrections to brightness and contrast.

I run a Zoom gathering of the Sonora Desert Astro Imagers every Tuesday morning. Part of this morning's gathering will be devoted to StarTools. Screen sharing on Zoom is actually a great way for astro imagers to exchange images and to illustrate the use of software in real time.

Thanks again,

Russ Carpenter

Thank you Russ,

ArcsinhStretch in particular is a solution to a problem that simply does not exist at all any more in StarTools; Luminance and Chrominance do not influence each other in StarTools, so a stretch that "preserves the original color" is obsolete.

Following the modules on the home screen left-to-right, top-to-bottom should give you a rough idea of how detail is progressively refined.

1. AutoDev to establish the global stretch
2. Contrast to improve visibility of detail in medium-to-large areas (e.g. most effective on for example image-filling nebulosity).
3. Sharp and/or HDR for digging out fine detail. Sharp in particular is very different from basic Wavelet sharpeners as found in PI; it throttles scale (controllable of course) based on noise grain evolution stats. E.g. it will bring out large scale structures if smaller scale enhancements would only exacerbate noise.
4. Deconvolution; note that Decon is completely unfazed by all the other local detail manipulation that preceded its invocation. As a matter of fact, it uses the knowledge how local detail was enhanced to its advantage during regularization.

All the above modules keep tabs on how signal (and its noise component) evolve. And the decisions you make in all the above modules, depend on the global visual stretch's context. That's why a new global stretch "tweak" would invalidate all of that visually, philosophically and mathematically.