Hi guys,
Ken - as you will have seen from my (DSI-IIC, CMYG) deBayer PJSR script, I used several methods to try and get my data 'colour perfect' at the individual image stage - long before they were aligned and integrated.
However, that script was conceived 'at the same time' as the introduction of the BackgroundNeutralisation and ColourEqualisation modules were being developed, and I would certainly, now, not even waste my time trying to compensate colours until well AFTER I had used ImageIntegration (with the stunning power of Winsorized Sigma Clipping for 'crud' removal) to sort out all of my data calibration, and I would then be invoking DynamicCrop and DBE (perhaps twice) before I felt that my master image was ready for the BackgroundNeutralisation and ColourEqualisation.
I don't know how well I have ever described the ColourCurve adjustment steps that I include in my script, but I figured out a process that just 'worked' - and stunningly well at that, given that my deBayer was of a CMYG 2x4 CFA matrix - which is a devilishly difficult beast to tame when it comes to extracting RGB from the RAW data!
Basically, what i did was to print out a very simple colour chart (included in a post elsewhere on the Forum, but whose link may still be broken pending Juan's final 'tidy up' needed subsequent to his server swap) - essentially three swatches of R, G and B, and three swatches of Cy, Mg and Gn - created as 'perfect colours' and printed out on 'proper' photographic paper by the likes of Costco, WalMart, etc. I then take a LONG EXPOSURE photograph of this (several actually) using EXACTLY the same steps (and equipment) as I would if it were an astronomical target - using natural daylight to illuminate the print, and a simple 50mm lens attached to whichever camera I am using. This will involve SIGNIFICANT 'workarounds' to filter the INTENSITY of the daylight, but NOT at the expense of the SPECTRUM of the incident light. You will have to work this out to suit your available equipment.
Once I have calibrated, aligned and stacked the RAW data, and deBayered where necessary, I have an image that has 'colour' - but not 'pure' colour.
Then I look at the 'Red' swatch, for example, and figure out where I need to place a 'modifying' point on the ColourCurve transformation to 'change' what 'should be' Red (and isn't) into 'actual' Red. I repeat this for Bu and Gn, and then add intermediate points for Cy, Mg and Ye. The LivePreview helps you to see what is happening as you prepare this horribly complex, twisty little curve. Then you apply the curve to the image, see what the result was - and consider repeating the exerciswe again - fine tuning, if you will.
Very importantly - you need to save ProcessIcons for each of the ColourCurve transformations - the idea being that these curves actually 'define' the 'colour response' of your camera, given the specific deBayer routine that you have used. My theory is that the camera has NO IDEA that it wasn't looking at a DSO - it just gathered photons, over a specific period of time. If the Calibration and deBayer process remains (effectively) 'constant', then the application of a 'ColourCurve' correction should also be a 'constant'
What you WILL need though - I forgot to mention this - is a swatch of 'Pure Black' and one of 'Pure White' (effectively the photographic paper itself, of course) in the image as well - to allow you to accurately define the Black and White points. Interestingly, you will notice that the 'greyscale' is NOT affected by your ColourCurve transformation.
You might be able to make sense of the process if you break down the post-deBayering steps that I have placed into my script.
Certainly, when I defined a double-correcting pair of ColourCurves from my 'simulated' colour test-card, and then applied the same Curve processes to RAW data acquired of a general countryside landscape, and of some highly colourful 'still-life' samples, the effect was incredible. And when I then applied the same process to ACTUAL data collected for M81 / M82 (using my DSI-IIC on a Moonfish ED80 APO) the effort had DEFINITELY been worthwhile.
I just haven't really made a big thing about it, because myself and Alex next door are probably the ONLY PixInsight users left who are still trying to image with CMYG one-shot CCDs - and even I have since turned back to my DSI-IIPro mono imager, to try and collect the best photons that I can (Alex may thus be the VERY LAST surviving DSI-IIC imager using PixInsight for post-processing - and he currently doesn't bother with Flats or FlatDarks, and relies entirely on Nebulosity's deBayer routine to create his 'master image', and then just throws very simple BackgroundNeutralisation and ColourEqualisation processes at the data as 'automatic' steps in his initial openeing gambit - and his data ends up quite nicely colour-balanced, thank-you very much - with the minimum of fuss, bother or effort).
It would be worth your while examining different strategies - one will suit you best, and you will know that it may need to be changed as your skill, and the power of PI, improve in the future.
Cheers,
for CC
Well I can hope