My PixInsight Tutorials: LRGB+Ha+OIII Comb. Workflow & NB Bicolour Workflow

[kayronjm]

Hello all. Here are two extensive workflow tutorials I've written with a large number of screenshots based on PixInsight workflows I go through:

1. PixInsight & Photoshop Workflow for HDR LRGB & H-a + O-III Combination Post-Processing: http://lightvortexastronomy.blogspot.com/2013/01/tutorial-pixinsight-photoshop-workflow.html
LRGB data is combined with H-a and O-III narrowband data to produce a realistic-looking colour image. Since Orion Nebula (M42) is used as the target, the workflow tutorial includes a HDR composite between two images of different exposures (1 minute and 7 minutes) so you can see all four trapezium stars in the core and the surrounding faint nebulosity. This HDR composite is done in Photoshop using layer masks but absolutely everything else of the workflow is in PixInsight.

2. PixInsight Workflow for Narrowband Bicolour Palette Image Post-Processing: http://lightvortexastronomy.blogspot.com/2013/01/tutorial-pixinsight-workflow-for.html
H-a and O-III narrowband data is used to make a Bicolour Palette image. Three different colour-combining examples are provided as one involves creating a synthetic Green with Steve Cannistra's method (applied to PixelMath) and the other two are different ways of combining the data (one looking more LRGB-like). The entire workflow is covered and is entirely based on PixInsight.

Hopefully someone gets something from them. I will be writing more as I learn more and as I do other things!

[pfile]

i read the first tutorial - nice. but i think you could save yourself a lot of time, and get a much less "burned" look at the core of M42 if you used HDRComposition to merge your exposures in PixInsight. this has to be done when the images are still linear. then you go do your stretches and to recover the core, you use HDRMultiscaleTransform to get the core details back.

rob

[cfranks]

Thank you for posting your workflow.  I have only scanned it so far but have learned something new already. 

Thanks again.
Charles

[kayronjm]

i read the first tutorial - nice. but i think you could save yourself a lot of time, and get a much less "burned" look at the core of M42 if you used HDRComposition to merge your exposures in PixInsight. this has to be done when the images are still linear. then you go do your stretches and to recover the core, you use HDRMultiscaleTransform to get the core details back.

rob

Possibly, though I've not had success with HDRComposition. There are extremely defined edges to the area that is replaced with the short exposure image. Is that what you mean HDRMultiscaleTransform fixes? I don't remember trying HDRMultiscaleTransform to circumvent the issue. It just seemed like HDRComposition was not doing its job as I expected it to. Mind you, the Photoshop layer mask HDR method takes all of 45 seconds to do!
It just seems long because it took me long to document with text and screenshots but in actuality, 45 seconds is all you need to do it. 

Thank you for posting your workflow.  I have only scanned it so far but have learned something new already. 

Thanks again.
Charles

You're very welcome indeed. Someone somewhere will benefit from all the writing and screenshots so it's worth having online!

[pfile]

HDRComposition was practically designed for the core of M42. HDRComposition does have some knobs - in particular the binarizing threshhold can be modified to change how the blending works. read the tooltips for the knobs in HDRComposition and you'll see.

since it's a numerical solution it's going to be superior to eyeballing where one image should blend into the other. almost every PS-blended M42 i see has some very strange darkening around the trap.

regarding HDRMultiscaleTransform, what i mean is that even after you've done the HDRComposition, when you stretch the data you'll end up with a blown-out looking core. but the data is still there; HDRMultiscaleTransform basically does a tone compression so that you can see it. attached is a 1200s exposure of M42, 5nm Ha, which has 300, 100, 60 and 10s exposures blended in using HDRComposition. the first screenshot is after the stretch, and the 2nd is after HDRMultiscaleTransform and LocalHistogramEqualization.

[kayronjm]

Thank you for that pfile, I will definitely give it a go and see how it works as I may have just done something wrong. I only ever tried using it once and it was a while back so it will probably work well if I try it again. Indeed a numerical solution is always superior to eye-balling something.

[kayronjm]

One thing I just realised I'm not sure how to go by at this stage is how to combine my long exposure Luminance with the HDR image generated by HDRComposition. I have to leave the data linear for HDRComposition to do its job so in theory I can only:

1. Combine the linear Luminance image with the long exposure RGB image whilst it's also still linear.
OR
2. Combine the stretched Luminance image with the stretched HDR RGB image generated by HDRComposition.

Number 1 destroys any and all colour saturation in the linear, long exposure RGB image and number 2 stops HDRMultiscaleTransform from later doing its job and revealing the core detail (as this detail has been removed by Luminance).

[pfile]

not sure i follow, don't you have a set of bracketed Lum exposures as well as bracketed RGB exposures?

[kayronjm]

not sure i follow, don't you have a set of bracketed Lum exposures as well as bracketed RGB exposures?

Sorry, I don't know what you mean specifically by bracketed (never heard that term in imaging before!). What I have is a Luminance exposure of 420s and then R, G and B exposures of 60s and 420s. I've colour-combined the RGB exposures into a 60s image and a 420s image with background gradients extracted, colours calibrated, etc. The 420s Luminance image is also background-extracted and noise on it has been reduced with help of a mask (to reduce it in low SNR areas only).

What I meant is that if I do an LRGBCombination of the linear 420s Luminance with the linear 420s RGB image, I destroy colour saturation and that has a significant effect once HDRComposition composites the 60s RGB with the new 420s LRGB and I stretch the histogram. If I just combine the Luminance with a HDRComposition image that only uses the 60s RGB and 420s RGB (not LRGB!) images, I destroy the detail in the core due to the Luminance being of 420s.

What I got to do last night was just do method 1 - LRGBCombination of linear 420s Luminance and linear 420s RGB and then to recover colour saturation, just perform a CurvesTransformation to Saturation only. Then I do the HDRComposition of this image with the linear 60s RGB and voila - colours are ok, Luminance is combined and HDR is achieved. I think that's the best bet. I still have to play around with the post-processing after this. I'm trying other people's workflows to see how they compare and if I can combine these with mine.

[pfile]

bracketing is a terrestrial photography term which means to change exposure lengths to capture highlight and shadow detail in the same scene. so it's what you are doing.

the only "right" way to solve this problem is to go back and get L exposures of varying lengths in order to make an HDR L that can be processed and blended with the RGB. another thing would be to try to create synthetic luminance images from the various bracketed RGB exposures and use those to do an HDR L.

but it sounds like you came up with a different way of doing it that works.

[kayronjm]

I see, ok, thank you for that! Indeed I also think the best way around the issue is to create a HDR Luminance image and a separate HDR RGB image and then do the LRGBCombination with both HDR images. I have raw 60s Luminance data but it's marginal as evidently my focus or the seeing wasn't that good. Avoiding using it allows me to see all four trapezium stars in the core, otherwise it's just one fat blob of whiteness!

Anyway, I successfully used HDRComposition and have done 100% of the post-processing in PixInsight, following some other workflows I've been looking into lately. This is what I came up with:

http://www.astrobin.com/30864/

That's a combined RGB+Ha+OIII (Ha enhances R and OIII enhances B).

EDIT: Changed link as the image I had yesterday wasn't very good after all. The colours were not very correct. I found out that this was due to the method of combining Luminance prior to stretching and then having to alter the colour saturation manually before a HDRComposition due to the LRGBCombination altering the colour saturation on my 420s image. In the new image, I've not used Luminance data at all. The process therefore HAS to be one where you create a HDR Luminance and a HDR RGB image and after both have gone through HDRComposition separately, you can combine them for LRGB.