A basic PixInsight workflow for LRGB images

[RickS]

I promised a few people that I would take my cryptic workflow mind map and make it into something more readable. The result is attached. I tried to keep it simple but still include some more advanced ideas. I hope it is helpful, especially to folks who have managed to tame some of the PI processes but are wondering how to connect them all together.

I don't claim this is the best way to do things but it mostly works for me. Each image will benefit from different treatment so don't treat this as something written in stone. Feel free to vary things and improvise. I usually do!

Cheers,
Rick.

[Spookyer]

thanks Rick, appreciate that. I may actually have some data soon, this will give me a start on how to deal with it.


Do you a narrow band workflow as well?

Brett

[ozstronomer]

Cheers Rick, much appreciated. Will try it out next time I have some RGB data

Geoff

[RickS]

I haven't written up a narrow band workflow, Brett, but you can use the same basic steps with a synthetic Luminance.

One simple way to generate a SynL is to do the LRGBCombination of the Ha, OIII and SII, get the colour about where you want it and then use ChannelExtraction to extract the CIE L* channel.

Another technique that has worked well for me is to use ImageIntegration (with no rejection) to combine the master Ha, OIII and SII frames into a SynL.



Hope it helps, Geoff!

[troypiggo]

G'day Rick, great work!!

Mostly lines up with what I've been doing (when I get to image!).

I notice you independently stretch the L and RGB, then just combine them with tweaking the lightness and saturation sliders. For me, and possibly/probably because I take L 1x1 but RGB 2x2, I have trouble getting the LRGBCombination looking right. All washed out, not matter how much I tweak the sliders.

Solution I found was to do the individual stretching of the L and RGB, but then throw in some steps where I extract the CIEL from the RGB, then LinearFit it to the L, then combine it back into the RGB before the LRGBCombine of L and RGB. Rarely need to touch the lightness sliders when doing that because they're already matched.

Other thing I do at the cropping stage is before applying a DynamicCrop to the first image, I make a Process Icon out of it so can just drag and drop in on all the masters (or use Image Containers).

[RickS]

Thanks, Troy. Small kids can get in the way of astro pursuits but they do grow up eventually :)



Yeah, I've seen that done but never needed it myself. I always bin x1 so I guess my L and colour subs are reasonably compatible. A lot of the time I don't even need the sliders.

I have used LinearFit to match narrow band integrations where there is a big variation in intensity between Ha, OII and SII.



That's definitely the smart way to do it. You can also create a process icon by dragging from the HistoryExplorer after doing the first crop.

Cheers,
Rick.

[troypiggo]

Didn't know about the HistoryExplorer creating Process Icons. Cool!

[multiweb]

Cool - thanks for that Rick. :thumbsup:

[gregbradley]

Thanks Rick.

Greg.

[pvelez]

Rick

that is super helpful - thanks for that

Pete

[allan gould]

Thanks for the write up Rick. Will be super helpful for me when the weather clears up and I get some decent data.
Allan

[RickS]

Thanks, chaps! If you have any feedback let me know and I'll update the document if needed.

[rogerg]

Thanks everyone (especially Rick) for your suggestions and workflows.

I'm struggling with enthusiasm for astronomy right now (watching endless clear blue sky days and nights go by) and I think it's largely because LRGB processing is just such a struggle for me still. Battling new workflow with CCDStack and/or PixInsight. I'll get there eventually.

[DavidNg]

Thanks Rick,

Any tips for the star mask? I've never got it right, always some small stars not covered as well as really big ones.

Regards

[Rod771]

This might help you David.

Click here for Pixinsight post (http://pixinsight.com/forum/index.php?topic=6001.msg40719#msg40 719)

[RickS]

There's no magic bullet unfortunately, David. Building a good star mask can be hard work.

Here are some of the things that I've found helpful:


if you have a colour image, extract the Lightness (CIE L*) and use that as the basis for the mask. If you have a mono image then just clone it.
if you're still working with a linear image then apply a stretch.
apply a heavy HDRMT dynamic range compression (as described in the PI Forum post that Rod mentioned.) Set the number of wavelet layers depending on the size of your largest stars and use several iterations (I usually try 4 or 6).
grab a small but representative preview and use it to test different StarMask parameters to save waiting time.
increase Noise Threshold if you're getting nebulosity or other non-star stuff in the mask. If you get really desperate there's always CloneStamp.
set Scale big enough to capture the largest stars you want to mask.
tweak the Structure Growth parameters to get the coverage you want and balance masking of small and large stars
you can use MorphologicalTransformation for fine tuning of the mask (or to gradually adjust a mask when making iterative changes like multiple applications of MT to reduce stars).
sometimes I do "fancy" stuff like: build a star mask, clone it, shrink the stars in one clone with MT Erosion, grow the stars in the other clone with MT Dilation, use PixelMath to "subtract" the little stars from the big stars making a donut mask that can be used to work on the outer edge of stars for fixing halos or reducing star size without dimming the cores too much.
don't forget to apply the masks to your image and flick between normal and inverted view to get a feel for how well they fit
sometimes a range mask (RangeSelection) can work better than a star mask.

Hope that helps... Experiment with the parameters and take your time.

Cheers,
Rick.

[RickS]

Hope you get your mojo back, Roger :) I quite enjoy the processing battle and I'm sure you will too once you climb the learning curve a little more!

[RobF]

Great stuff Rick.

Interesting you do gentle noise reduction first thing on the Lum master. I wouldn't normally do any NR until the end depending on how good the data was and how hard it had to be stretched.

Anything like this that gets you thinking is good as no "right" way of course.

[RickS]

Thanks, Rob.

I think I first saw an example of noise reduction of linear data in a post by Juan introducing noise reduction using ATWT. It seems intuitively, to me at least, a good idea to tackle noise early and it does allow you to start with a slightly more aggressive stretch.

I use a clipped luminance mask to protect the bright areas and normally prefer ATWT or MMT with noise reduction only at the smallest wavelet scales. I work on a preview at 1:1 or higher zoom to check the effect and always err on the side of caution. It's really just a small scale smoothing of the background areas rather than full-on NR.

The thing I really can't decide is whether it is better to do this NR before or after deconvolution. I haven't been able to detect much difference so maybe it doesn't matter?

Cheers,
Rick.

[RobF]

I just know I'm "bad" and always jump in the deep end stretching, teasing, pushing. Then way down the road often end up going back to look at deconv or NR if the data seems to be looking promising.

Mmmm. Sounds like my cooking too actually :P

[rogerg]

To create a preview do I really need to write down the pixel top, left, width, height values or is there some way to select an area of an image to be the preview?

Edit: to clarify: I am going Preview -> New Preview. Other preview options ("New Preview Mode" etc) are greyed out/not available for some unknown reason ...

[DJT]

;)

Hi Roger
Click Alt N then you can define the area you want to preview using your mouse.Once you have defined the area, a new preview tab is created in your workspace.

If you want the whole image to be a preview, a quick way is to drag the descriptive tab of the image down below the original Within your workspace and then Let go.

Preview options are greyed out because they only seem to work on preview tabs themselves. If you right click on your preview tab once created you can also get to the options that way rather than through the menu.


This pixinsite stuff is pretty good but takes a bit of getting used to. I learnt the preview stuff just last night using the IP4AP tutorials :)



Cheers

[RickS]

Yep, the quick way to click and drag a preview is the New Preview Mode command: Alt-N as David suggested or the equivalent menu bar icon.

There are a couple of ways to create a new preview using numeric position and size - Preview>New Preview from the menu and the "preview" command in the process console command line.

You can also copy an existing preview to the same window or another window. Select the preview, click on the preview name and drag/drop it on the left hand window bar. I find this really useful for comparing different processing parameters.

Cheers,
Rick.

[rogerg]

For some reason Alt-N and the preview menu options are greyed out and unavailable for me. Any idea what criteria make creating a preview available?

The only preview related menu/quick key combo available is the Preview -> New Preview which requires numbers typed in :shrug:

[RickS]

The only time I have noticed the New Preview Mode command icon greyed out is if you don't have a suitable image selected, e.g. if no images are open or an existing preview is selected.

[rogerg]

Hmm, my suspicion also, but ...

1) LRGBCombination ... which creates a resulting image.
2) Save resulting image as .FIT
3) close all files
4) open LRGB .FIT file
... no preview options available :shrug:

[RickS]

Roger,

Some people have reported problems with Alt-N not working reliably on OSX. If you're using Alt-N have you tried the New Preview Mode icon instead? View>Tool Bars>Mode to display the group of icons that include it.

Cheers,
Rick.

[rogerg]

Toolbar icons, and menu options are greyed out. Try clicking them anyway and nothing happens, and no evidence of them accepting the click.

Ok, seems we've tried the obvious "how it should be done" so I'll mess around more tonight with more different formats of file and the like, see if I can find any reason the preview functionality isn't working.

Computer is a PC, a recent and well spec'd 64bit one.

Thanks.

[RickS]

Might be worth a post on the PI Forum if you run out of ideas: http://pixinsight.com/titles/pixinsight_forum_title.png

[Octane]

Apologies for the thread necromancy.

Roger, are you still having this issue?

I was having the preview (Alt/Option+N) issue on my MacBook Pro.

I thought I'd try and find out what was causing it.

I hit the Caps Lock key once, and, then, it started working. You can then disable Caps Lock, and, it should work as per normal.

Small glitch.

H

[rogerg]

Thanks H.

I ended up solving it by a PC reboot. Not sure exactly what the root cause was.

Since then I gave ended up sticking to using CCD Stacker for LRGB combining just because it is simple and straight forward.

Roger.

[Octane]

PixInsight just looks scary. It's pretty intuitive after a little while. And, surely, the results are worth it. :)

H

[rogerg]

I find it good for various things, like reduction/calibration of CR2's, aligning of CR2's, and DBE. But still find it more fiddly than others for other tasks.

I quite like CCDStack for a few reasons - the most notable is probably it's visual representation of sigma clipping. Being able to preview what data is being rejected very quickly and easily rather than having to do lots of trial and error with sigma values and waiting for image proessing in PixInsight.

There's pro's & cons.

Intuitive certainly isn't a word I'd associate with PixInsight, with it's unique use of buttons/icons, problems working with dual screen, and other usability issues :)

[RickS]

It takes longer to do a careful integration in PI but it's the only amateur package I've used that shows you SNR measurements and let's you maximise SNR by varying algorithms and rejection parameters. Not good for instant gratification but nice to know you're getting the best possible result from the data.



That's partly the price that has to be paid for cross platform support. I quite like the PI UI but I have used some shockers :lol:

Cheers,
Rick.