›

How to get True Colors?

Gary Schnettler
in PixInsight
Hello.  I know that color can be tricky in astrophotography, but I'd like to know at least if my pictures have true color or modified color to make them look better/more artistic. 

For a regular RGB photo, stacked in Pixinsight, normally if I click the nuke button in PI right after WBPP it'll start off looking very bad, such as excessively green or blue. Then I click Unlink in STF and click the nuke again and then it typically looks pretty good.  Then I bring up SPCC, carefully set my camera and filters and run it.  Now, I have a perfectly calibrated color image, right?  Or do I?  What does that even mean?

I was just watching the Adam Block M83 workflow videos where he ran SPCC, and then said there was still too much green so he ran SCNR at around 50%.  Later he used STF to reduce the green even more.  He zoomed in on the background and noticed some color there so he adjusted the colors in STF again and applied them by dragging the triangle to Hist Trans.  He also brought up Curves and clicked on the lowercase b and then made some adjustments to increase the blue and to shift the red to yellow. 

After all of that his M83 colors looked very good. But, were they accurate?  Some of that seems necessary.  For example, space should be black so if there's color we need to remove it, right?  

To sum up my question:  Does SPCC give your image true colors, and if not, what does it do and how do you actually get an image where the colors are as close to reality as possible?

Comments

  • BlockHeadBlockHead
    edited December 2024
    Gary,

    I will answer your question. However, you have to do a little work first.
    You need to watch my entire video that explains what color calibration is and how SPCC is doing this- more accurately than any other tool currently available.

    Zero people other than me have taken time to fully explain and fundamentally answer your question

    See:

    Then, after reviewing the video what I say next will be helpful. If you want calibrated color with a reasonable choice of color balance what SPCC gives you is about as close as you can get to a consistent and accurate color (calibration and balance). 

    However, my adjustment of green noise is part of an electronic (instrumental) signature I feel I am compensating for. This is a choice. The adjustment (enhancement) of blues/yellows through Lab color space adjustments is just that- an artistic choice. It is a move away from the perfectly calibrated data I obtained after SPCC. 

    So, if you want to be like James Wray:
    In which  is he produced every image in precisely the same way- that is an option.
    You would be what many would call a purist.
    (PI Developers would laud you.)
    I am not a purist. But I do have constraints. My choices are ones that come from experience and an understanding of the instrumentation, the data, and an artistic sensibility that communicates the interesting elements of an subject. 

    Most imagers are far, far, far away from this and screw with everything with abandon. I find the lack of consistency or logic behind choices just as bad as a poorly executed image. 

    Keep in mind that virtually no professional observatories use broadband filters as we do. They are all missing colors we incorporate. So you cannot look to those images as exemplars. 

    To sum up SPCC gives you very precise color calibration and therefore a faithful color balance (of whatever the chosen white point is). You need to parse what you mean by "reality". In the language of image processing this is color calibration and color balance. 

    -the Blockhead
  • Gary Schnettler
    I watched the SPCC video.  Thank you for making that video - it was very good!  I love the science behind it.  But now I'm even more convinced that once the color is set by SPCC, then the color should be good and we should leave it (if we want pictures with accurate color). 

    I think that leaves 4 problems to deal with:

    1) Stretching apparently washes out color so you need to bump up the saturation (this was mentioned in one of the GHS videos too).  Maybe this is ok if it's done evenly by simply raising the color saturation slider as is done in the M83 video.  Maybe not - see #4 below. 

    2) If SPCC is pretty much perfect, why do we need to do SCNR and remove green noise afterwards?  It seems that if there is extra green that SPCC should take care of it. 

    3) If SPCC is pretty much perfect shouldn't we be left with a perfectly black background?  If not, and we have to adjust it then that seems like doing so would throw off the perfect SPCC colors in the galaxy or nebula. 

    4) The M83 video shows that RGB colors have peak saturation at different brightness levels (see the attached photo and the M83 Luminance video starting about 1:05 in) where it recommends going for about a 0.8 brightness on the Luminance so that the colors aren't washed out.  But, it seems to me that the 0.8 is somewhat arbitrary and, since the colors don't have peak saturation at the same points, then if we chose a 0.7 or 0.6 brightness target for the Luminance then our color balance would be different. Perhaps this is why some people skip the L and just do RGB imaging. I'm really torn on this though since the M83 LRGB video made the whole Super Luminance process look amazing. 

    I'm not doing science with my images at this point in time and therefore it's just art so I guess it doesn't matter.  But when someone comments on my pictures and asks if the colors are true or altered I'd like to be able to say that they are the true colors, calibrated very accurately using GAIA data from the European Space Agency. 

    p.s. I ran into the example mentioned in the SPCC video where the stars that are supposed to be in a straight line aren't.  I was puzzled at first but then figured out that it was due to having "Sony" filters in SPCC instead of Chroma. See attached images of SPCC with the incorrect and then correct filters.

    image


    imageimage
    RGB Screenshot 2024-12-03 at 1.05.52 PM.png
    774 x 772 - 249K
    IMG_5901.PNG
    1026 x 1724 - 241K
    IMG_7678.png
    1030 x 1938 - 285K
  • BlockHeadBlockHead
    Gary,

    1. Once you move into the non-linear domain- no matter how careful you are, you will be affecting the color balance you achieved with SPCC. However this can be small if you are careful to make modest stretches and raise the color saturation. The idea with color science in terms of color space are very complicated. The calibration we did before is numerical- and independent of how we display or stretch the data. If you are going to make a measurement- the color calibration and then the resulting value is what you want. For making a pretty picture- we used a color calibrated and white-balanced image as are starting point. We are trying to remove the bias  or systematic errors caused by your camera and telescope. That is the idea behind calibration. You don't want your images to be colored by your camera because it is more red sensitive and makes images redder than they would be for an equal response at all wavelengths right?

    2.  No. The calibration is of the signal...the flux. The background is mostly the noise. The calibration is mostly a scaling/multiplicative adjustment. The background is an offset. Just think about this stuff deeply Gary. If you are taking a picture when the moon is in the sky- there will be "blue offset" to all of your images because the sky is blue. When you apply SPCC...this will color calibrate based on star flux (of which the moonlight does not contribute much because the stars are much brighter than the moon and because when the PSF is measured the background next to the star is subtracted). So the white balance is based on the stars. This is why SPCC has to perform a second step of Background neutralization. The offset is not part of the color calibration directly. Now... I really take offense to the sentiment "If SPCC is so good/perfect.."  I am doing my best to answer your questions. Neither the color calibration or the offset will be perfect. I never said anything was perfect. The noisy characteristic of cameras can still be biased in color. Our sensors are generally peaking in sensitivity in the green...so there will be a greater standard deviation of green-colored noise (and perhaps the sky is brighter here leading to more poisson fluctuations). So the SCNR adjustment I did was not a color balance adjustment..it really isn't affecting the overall color of the brightest pixels (stars/galaxies/nebula)- it does more strongly affect the noisiest values- which is the background. The background independently colors your faintest signal. You take a picture of the Orion nebula... which is red and blue...but the faintest parts have excess green because of the background. You don't want this background instrument/sky color to be controlling your result do you?

    4. Yes, 0.8 is arbitrary. What number would you pick? The saturation level and all of that I mention in point #1 above is the issue. This is a color space representation problem of non-linear data. I am the first astrophotographer to clearly communicate the issue and suggest a value that is practical (perhaps in 2003 or earlier). It is interesting that I am typically asked for a recipe..a step-by-step outline of what to do. Usually I answer in more general terms to be flexible in terms of processing. But here I am..you got me, I am nailed down to a value that I find practical from experience- and oh boy...seems kinda arbitrary Adam..doesn't it? I can't win. :)  Let me add that the Hue and Saturation changes in a particular color space are complex. Your simplified statement that "the color balance would be different" ... is not complete. You can still retain a semblance of color balance even when changing the saturation... even if this does shift the Hue a little. 
    Doing RGB processing does not relieve you of exactly the same issues. Screen representations of a particular color space always have a "lightness" ... which is correlated with the luminance of LRGB (or Lab). I really am not an expert in color spaces. There is plenty of information on the internet (wikipedia) on this.  

    Yes, the fits will not as accurate if you are not inputting the actual filters being used. 

    -the Blockhead
  • Peter Rejto
    Adam, I really appreciate that Gary raised this issue if only that it provoked you into this detailed and informative response. In my view there is a lot of valuable information and reasoning in your reply that we (I) should pay attention to. Thank you!
    Peter
  • BlockHeadBlockHead
    Thanks Peter I appreciate that. I spent some heartbeats writing the above.
    -the Blockhead
Sign In or Register to comment.