StarTools

It would be nice to have a Photometric color calibration option as in pixinsight.

This is useful in areas with clear stars such as galaxies, clusters etc.

This could provide an accurate and easy way to calibrate stars using plate solving and then the actual star colors. This is specifically useful for RGB broadband filters.

Thanks, Eric

Its also in Siril. I often find my results with photometric color cal to be excellent. Will be following this thread to see what Ivo has to say

PCC's usefulness is extremely limited (I would personally argue - detrimental to a proper understanding of one's color rendition). Intuitively it may make sense to "just" measure objects in your image against known spectra. In reality, this is fraught with inaccuracies and (necessary!) arbitrary assumptions. There is literally nothing that is special about PCC that cannot be achieved in StarTools Color module (and in a much more considered way). At the same time, PCC gets just as confused by things like chromatic aberration, coma and channel alignment issues as StarTools' "default" color calibration routine.

User wimvb on the SGL forums, for example, recently demonstrated it quite well using PI's PCC;
As wimvb wrote;

The tag in each panel indicates the white reference that I used in colour calibration:

Star spectral types: (from red to blue) O5V, B0I, A0I, F0I, G0I, K0II, M0III, O5V and Photon Equalization

Galaxy types: ASP (Average Spiral Galaxy, the default in PCC), Sa, S0, Sb

"Take your pick"

(his post is here, though please be aware there is a staggering amount of misinformation in that thread, mostly by user vlaiv).

They are all valid renditions, but it is likely many users of PCC have no clue how or why that is so, or how these renditions came about, or - crucially - which one to pick.

E.g. if it weren't clear yet, there is no such thing as "actual star colors"! The road from photometry to color on a screen is full of arbitrary assumptions (one of the most important being the choice of a white reference) and points of failure/error. You can read more about the vast amount of arbitrary assumptions and factors that influence the final coloring of your image here, noting that this is just for terrestrial photography, whereas astrophotography adds a slew of other factors, assumptions and variables.

Sometimes I wish PCC would go away - it gives many beginners a false sense of security and makes them think less (if at all) about their color balancing. It also has the nasty side effect of making datasets noisier (if a synthetic luminance dataset is not saved away first), again putting newbies at a possible disadvantage.

The amount of times I have seen PCC applied in Siril with a subsequent green dominance neutralisation (because, hey, PCC's results must be correct, right? so any left-over green is bad) makes me really, really sad.

I hope this explains why I am not in any particular hurry to implement something along the lines of PCC.

Ivo,

First of all, stars do have actual colors and spectra which have been measured and are in databases.

I agree with you that setting color in Astro photos is a challenge.

PCC when used properly, can often give very appealing results which are based on a quantitative method.

For example, if you use the galaxy settings for galaxies, and the G0V setting for stars, this can work well , but is not foolproof. It is not typically useful in nebula regions.

The color module in star tools is appealing, but does have a subjective quality to the results… it seems to rely on user tweaks for large potential color shifts.

PCC would involve a certain amount of Coding, but might add an alternative more quantitative choice.

Eric

EG1 wrote: ↑Mon Aug 30, 2021 12:55 pm First of all, stars do have actual colors

I urge you to peruse the links I gave you - you will come to understand that this statement is not true.
Color perception (and rendition) is highly subjective and only the result of a great many arbitrary transformations and contexts. It is a deep, deep rabbit hole. "It's not that simple" doesn't even begin to cover it!

and spectra which have been measured and are in databases.
I agree with you that setting color in Astro photos is a challenge.

Indeed!

PCC when used properly, can often give very appealing results which are based on a quantitative method.
For example, if you use the galaxy settings for galaxies, and the G0V setting for stars, this can work well , but is not foolproof. It is not typically useful in nebula regions.

You are absolutely right that, for starters, the choice of a whitepoint (as you allude to, for example a ~5800K G2V star like our Sun) is indeed entirely arbitrary. Not to mention that PCC will yield different results, depending on

• filter response of your Bayer matrix or filter wheel (red filters, for example lack a violet bump in the blue part of the spectrum, whereas this bump is present in DSLRs, as humans are able to perceive violet as purple). EDIT: note also that most Bayer matrices and RGB filter sets do not correspond at all well to the spectral response of the photometry filters (like the set used for APASS).
• the presence of chromatic aberration
• and/or alignment errors
• and/or atmospheric extinction
• and/or smoke haze
• and/or light pollution gradient remnants, sky glow
• and/or slight uncorrected flat frame issues, etc.

Not to mention that coloring then often gets mangled (in other applications) by processing luminance and chrominance entangled rather than separated, which then necessitates saturation tweaks, which then require color space translations with gamut limitations, etc. That and the distorting effects on color. of local dynamic range manipulations (see further down re:psychovisual context preservation).

I cannot stress enough that PCC is not a silver bullet or will somehow yield anything that is canonical or "closer-to-the-truth" than other methods.

The color module in star tools is appealing, but does have a subjective quality to the results… it seems to rely on user tweaks for large potential color shifts.

On the contrary; StarTools' Color module goes to great lengths to offer deep, useful (rather than "feel good") control over coloring. This includes renderings that are more scientifically valuable (ratio preservation across different color spaces) and renderings that are psychovisually consistent/correct (if you so choose of course).

For example, due to the separation of luminance and chrominance early on, luminance does not impact coloring psychovisually. You may (or may not) have noticed for example that galaxy dust lanes almost never appear orange in StarTools. That is because StarTools respects and attempts to preserve the psychovisual context (within the limitations of the gamut of course) that makes the same RGB value appear brown or red in one context and orange in another (see also this video).

I know that Juan/PI also passionately believes in picking a color balance that conveys the most discernible coloring (artifacts like chromatic aberration not withstanding of course) - it is the most informative rendering and therefore the one with the most scientific value. This practice is precisely what the Color module implements to come up with its default coloring (e.g. a variant of a "grey world" sampling algorithm that uses luminous objects to calibrate the white point against, so that they all maximally stand out, closely mimicking human color constancy perception)

I hope from the information provided that it is clear that the results from PCC are no more correct or useful in any way. Relying on PCC for color calibration alone adds a point of failure to your workflow (a failure mode I have seen one too many times when used by newbies). PCC can not (and should not) be more than a ballpark starting point. If that.

Of course, if you so wish, you can use the balancing factors PCC comes up with in StarTools by setting the Bias Increase and Bias Reduce parameters to those factors - if you consider this the best of both worlds.