PixInsight journey

[gregbradley]

I have started on learning PixInsight in earnest. I have the Astrophotography Manual book by Chris Woodhouse which I bought for its section on PI. I also have the subscription to IP4AP PI tutorials.

Up to now I have only really played with PI and used dynamic background extraction to clean up some gradients and also morphological transformation to round out some elongated stars.

So I am a newbie to the software so I can post simplified how to's in a thread here if anyone is interested and we can learn together. I would take the "look at me I am so smart I use non comprehensible big words to show how smart I am" out of it. PI is riddled with that. I guess morhphological transformation (morphological means to do with the shape/form of something, transformation means to change into something else) sounds a lot smarter than "star rounder tool"!

I started with the tour of the various icons on the windows which I found to be the most confusing thing about PI. Microsoft windows icons we are all familiar with but what are those extra symbols on the windows and which one should you be using etc???

Anyone let me know if there is interest and we can start from total newbie and work our way up to a finished image. A lot of the tools are not that hard but getting an image calibrated, registered, data rejected and combined like in CCDStack is new.

Greg.

[RickS (Rick)]

Originally Posted by gregbradley

I would take the "look at me I am so smart I use non comprehensible big words to show how smart I am" out of it. PI is riddled with that. I guess morhphological transformation (morphological means to do with the shape/form of something, transformation means to change into something else) sounds a lot smarter than "star rounder tool"!

The guys who write PI are astronomers and computer scientists so they use the technical nomenclature they are used to. I don't think it's an intentional elitist thing. That "star rounder" is actually a generic shape changer and the name reflects that.

Originally Posted by gregbradley

I started with the tour of the various icons on the windows which I found to be the most confusing thing about PI. Microsoft windows icons we are all familiar with but what are those extra symbols on the windows and which one should you be using etc???

Agree that the non-native interface is a PITA to start with. It's a consequence of making PI portable to operating systems other than Windows (how many other applications do you know that run on Windows, OS X, Linux and BSD Unix?) The Photoshop GUI is similarly unlike anything else in Windows but we forget that because we're used to it.

Anyway, I think your idea for the thread is good, Greg, and I'm happy to chime in from time to time if you want. If not, I'll just lurk and bite my tongue

Cheers,
Rick.

[gregbradley]

Originally Posted by RickS

The guys who write PI are astronomers and computer scientists so they use the technical nomenclature they are used to. I don't think it's an intentional elitist thing. That "star rounder" is actually a generic shape changer and the name reflects that.


Its more about being able to communicate clearly to the user. Its not their strong point and often isn't in the software world.

It is a difficult problem as you need to create new words to match their tools yet stay in communication with their users. Some do it better than others but I'd say here its quite a barrier to their program being used universally.


Agree that the non-native interface is a PITA to start with. It's a consequence of making PI portable to operating systems other than Windows (how many other applications do you know that run on Windows, OS X, Linux and BSD Unix?) The Photoshop GUI is similarly unlike anything else in Windows but we forget that because we're used to it.

I am not being critical of their interface, rather its simply different and is a barrier in itself as clicking those little symbols does not really reveal what they do. PS has sliders and a simple checkbox for preview so you can muck around with it and get some sort of a feel for it easily. As you say having it work in different operating systems no doubt is a feat of computer engineering. But surely the vast bulk of astrophotographers are Windows based or perhaps some are Mac users. Limux? Would that be a professional observatory or someone who uses Linux in their career?


Anyway, I think your idea for the thread is good, Greg, and I'm happy to chime in from time to time if you want. If not, I'll just lurk and bite my tongue



Cheers,
Rick.

Of course you are welcome as our local PI expert and all round good guy!

Greg.

[loc46south (Geoffrey)]

Hi Greg - I found the easiest way to get to grips with Pixinsight was with a Laptop and your Desktop. There is a wealth of info in Video form on the net - Harry's Astroshed is your first stop. Lock yourself away in a little dark room and run the videos on the laptop while doing the processing steps on your desktop. At least that way you can pick up the basics fairly quickly.

I very much doubt if you will ever become really expert at it, the software at times seems to evolve faster than normal mortals can comprehend it.

Good Luck
Geof Wingham

[troypiggo (Troy)]

Everyone points to Harry's videos as a good starting point. I'd like to also point to http://lightvortexastronomy.blogspot...tutorials.html for some excellent, thorough, well-written tutorials that are regularly updated and added to.

[peter_4059 (Peter)]

Good to see people getting on board, the workflow isn't too hard:
Register, stack etc
Use decon or other sharpening tools on the lum then stretch lum and RGB with masked stretch or histo trans.
Only apply noise reduction (TGV) to high signal areas by applying masks first
Key is trial and error so use duplicates and previews to try settings before committing changes.

Enjoy!

[Ryderscope (Rodney)]

What I have done Greg to assist with the learning journey is to start a spreadsheet in which I have a separate tab for each processing job that I undertake. I then record in detail each processing step that I make and the parameters that I have used (at least most of the time anyway!). In this way I can go back and check what I have done and hopefully repeat what works and lose what does not.
R

[Occulta (Chris)]

All

Sky & Tel are running a PI Webinar hosted by Warren Keller on November 11.

Details here:
http://tinyurl.com/nepgctz

Chris

[gregbradley]

Originally Posted by troypiggo

Everyone points to Harry's videos as a good starting point. I'd like to also point to http://lightvortexastronomy.blogspot...tutorials.html for some excellent, thorough, well-written tutorials that are regularly updated and added to.

Thanks for those links. I like the lightvortex one. Sometimes video tutorials can move a little slowly and then too fast when you want to practice something was just gone over.

I learnt a lot about the way you handle the basic operating system in less than an hour there.

I can see straight away the basic operating system is very powerfully an intelligently organised. Good stuff!

I can post comparisons of tools in Photoshop and CCDstack to PixInsight as I go along.

One I can mention already is automatic background extraction tool. Simply clicking on the gradient affected areas and setting it to subtract does a very nice correction of background gradients. You may need to do it twice. It seems better than Photoshop based gradient corrections like GradientXterminator or apply image subtractions of dust scratches/blurred images with the main object healing brushed out to create a flat like image of the gradients and setting an offset so it doesn't go too dark.

Morphological transformation can round out stars better than any PS tool (of which there are really none, perhaps Star Tools does to some extent).

Early days for me to tell, but it seems registration tools may do a better job than CCDstack does. CCDstack's CCDIS plugin for registering images can fail occasionally. It did a good job on some DSLR images recently that CCDstack failed to register.

Greg.

[Regulus (Trevor)]

morphological transformation (morphological means to do with the shape/form of something, transformation means to change into something else) sounds a lot smarter than "star rounder tool"!

- Made me laugh

[troypiggo (Troy)]

Here's my "go to" basic workflow for LRGB. It's no-frills, no sharpening or deconvolution, no noise reduction etc - just bare-bones calibrating, registering, stacking, channel combining, white balance, and stretching. Gets you an LRGB image to look at, then decide what you want to tweak or improve for round 2.

1. Load all files (bias, dark, flat, light) into the Batch Preprocessing script. I use master bias and darks, but flats and lights are per session. Typically I leave the bias/darks/flats tabs default settings. In the lights tab I don't have "calibrate only" ticked, I do use Cosmetic Correction via a process icon, I don't drizzle, and I apply Image Integration. When this completes, you will have individual master LRGB files.
2. Use LRGBCombination process to combine the RGB only, with all other settings default.
3. Use DynamicCrop to trim the perimeter of images where stacking leaves that funky border. The trick here is to draw the crop rectangle in the RGB window, but don't execute the process right away. Drag the blue triangle onto the desktop to create a process icon that remembers the crop geometry. Now you can execute the crop on the RGB image. Close the DynamicCrop process. Now drag and drop the process icon you created on to the L image. Should be identical geometry to the RGB.
4. Create a preview on a background part of the RGB image. This is for background colour balance, so pick somewhere without nebulosity. Don't think it matters if there's a few stars in there.
5. Open the BackgroundNeutralization process. Leave defaults but tick "Region of Interest" and choose the preview you created above. Run that.
6. Open the ColorCalibration process. All defaults, for background reference, choose ROI and use the same preview as 5 above. For white reference, choose structure detection typically with default values. Run it.
7. Do a MaskedStretch process with defaults on both the RGB and L.
8. Using ChannelExtraction and CIE L*a*b* extract L from the RGB image. It should create a new file called RGB_L.
9. Open LinearFit and using your original L as the reference image, drag the blue triangle onto the RGB_L image.
10. Now ChannelCombination the RGB_L back into the RGB using CIE L*a*b*.
11. Finally, LRGBCombination your original L into the RGB.

[Octane (Humayun)]

And, then, there is mosaics...

H

[RobF (Rob)]

This guys workflow worth a read too.
Seems to know a bit about it.
http://www.iceinspace.com.au/forum/s...d.php?t=117544

[troypiggo (Troy)]

Originally Posted by RobF

This guys workflow worth a read too.
Seems to know a bit about it.
http://www.iceinspace.com.au/forum/s...d.php?t=117544

Nah. He's all talk

[RickS (Rick)]

Originally Posted by troypiggo

Nah. He's all talk

Yup, and it needs to be updated a little. I'll try to do it some time soon.

[gregbradley]

Thanks Troy and Rick for your contributions. I'll certainly try out both of your workflows and they mirror what I am doing in CCstack and Photoshop currently.

I did a test last night of CCDstack versus PI registering and stacking some Ha subs.

Using the standard settings of PI Batch Preprocessing Script which is set to some data rejection, winsorised sigma clipping stacking and average combine I checked the values of the resulting master from both programs.

The half flux measure of stars was a lot higher with PI. The FWHM was slightly smaller with CCDstack (3.7 versus 3.8 so not very meaningful) but the difference in intensity was like 55,000 versus 41,000. So brightnesses are being better preserved in PI.

More tweaking with parameters as Rick has alluded would have to help.

Timewise both programs took about the same amount of effort. PI looks like it would be easier to set up for repetitive processing but CCDstack has an automated routine now but more complex to setup.

Also one minor but annoying thing about CCDstack is it does not number the subs when you open a stack. PI does. Yaaaahhhhh! I don't know how many times I have had to count the number of subs in a stack to keep track of the total exposure time of the image I have taken!

I used Lanczos256 and Bicubic Spline 36 in CCDstack and got the same FWHM values which surprised me. I though Bicubic blurred slightly.Lanczos256 takes ages though to process a stack so you have to be keen.

By the way for those taking the plunge the lightvortex site is best so far. I read the intro tutorials for only an hour and got much more proficient with the PI icons and operating system. Once past that I think the processes are not really that different to using something in Photoshop in terms of difficulty.

Speaking of that - anyone wish to share their setup process icons? I believe you can do that with PI which is a bit like importing premade actions in Photoshop (Noel Carboni actions for example). A lot of basic settings must be similar image to image with variations of course but routinely it must be often similar.

Greg.

[lazjen (Chris)]

Greg, you'll have to update your profile to change from "Registered User" to join the PI Cult. see above.

You've probably already discovered it, however the PixInsight forums can be useful too: http://pixinsight.com/forum/index.php

[sil (Steve)]

I'd just like to add the talk here seems to follow the same assumption that afflicts most astro imagining software: that everyone is capturing data the same way and focus on workflows that might not be of any help. I have an SBIG I've never used since I had my stroke soon after I bought it so I do what I can with what I've got. Likewise many people have to try AP with what they have or can afford.

I use differing workflows if I'm processing video or still image captures. Also if i'm imaging the sun/moon, planets, comets, nebulosity, galaxies I do things differently. So I started with understanding the simple workflows to know what and why they were doing. I took notes on how to do each step and worked on refining each little step to be optimal. So I've got a set of instructions I can chop and change as i learn more or as try capturing different ways.

For example i used to start with registration as step one. Now I sometimes start with PIPP to turn video into stills weeding out bad frames or I use PI script to keep the best FWHM value frames from a set of stills. So I start with trying to weed out bad frames before registration. Of course it really starts with your capture method too.

So my primary astrophotography these days is a DSLR on a tripod, no scopes, no tracking etc so I my workflow centers on using hundreds of images and I funnel my video captures to images (as lossless as possible) so I can use the step from my notes that I need when I need it. As I learn new techniques for doing things like "rounding stars" or "enhancing faint structure" I can just slot that in when i need/want it.

For newbies I'd say learn how to register (align) images, integrate (stack) images and stretch (adjust levels). That will get you an image, everything else is asking "whats wrong with the image and how can I fix it?"

With PixInsight I always keep my first integration image intact and save next steps to new files. Then if I want to try different things I can just start with the integrated image. If I want to do something different in registration or integration I always have my sources images intact and can rebuild from there. Be aware there are multiple ways of achieving the same thing (like reduce noise, round stars) that may not be obvious from tool/menu names. So when you follow a tutorial take note of the tools being used and what they are doing to an image so you have options to try for a step. For example you will treat images differently if you are doing photometry (taking measurements from the signal) from aesthetically (just making it pretty).

[gregbradley]

Originally Posted by lazjen

Greg, you'll have to update your profile to change from "Registered User" to join the PI Cult. see above.

You've probably already discovered it, however the PixInsight forums can be useful too: http://pixinsight.com/forum/index.php

Hehe. PI is a little bit arcane.

Greg.

[marc4darkskies (Marcus)]

Originally Posted by gregbradley

Hehe. PI is a little bit arcane.

Greg.

How to place your shoes in a shoe box – A PI workflow:

1. Buy three university degrees - numerical analysis, logic and computation and astronomy for $300,000.
2. It may also help to have studied Best Practices for Application User Interface design.
3. Make a virtual shoebox using the FGVC (Fabricate Generic Virtual Container) tool. Set the material setting to Cardboard but leave the other 28 parameters at default settings.
4. Make a virtual shoebox lid using the FGVCC (Fabricate Generic Virtual Container Coverage) tool. Again set the material setting to Cardboard and leave the other 77 parameters at default settings.
5. Overlay the virtual box with the virtual lid using the VCIV (Virtual Container Item Verification) tool
6. Use the verified design to manufacture a real shoebox by the normal method (don't worry, this is pretty intuitive)
7. Remove the lid from the shoebox using the NVCCD (Non-Virtual Container Cover Displacement) tool. Thankfully this only requires 11 parameters
8. Place your right foot into the shoebox to make sure it will fit
9. Step out of the shoebox
10. Find a toolbox
11. Resize the shoebox to fit perfectly to your foot as required using the DGCMT (Dynamic Generic Container Morphological Transformation) tool
12. Resize the lid to the shoebox as required using the AGCCA tool (Adaptive Generic Container Coverage Alteration tool)
13. Put the lid back on the shoe box – no tools required … duh!!!
14. Repeat the sizing method for the left foot using some of the above steps. NOTE in step 11 - do NOT reduce the size of the box
15. Stand on your head
16. Remove the lid from the shoebox with your eyes closed and place it next to your left ear
17. Rummage around in the toolbox for the ASD tool (Adaptive String Deconvolution). Don't worry, even though you can't see, it will feel intuitively like an ASD tool
18. Undo the shoelace of your left shoe with the ASD tool in your right hand
19. Remove the left shoe with your left hand
20. Place the left shoe into the shoe box using your left foot and balance it there
21. Now undo the shoelace of your right shoe with the ASD tool in your right hand
22. Remove the right shoe with your left hand and balance it in the shoebox with your right foot
23. You should now have both shoes in the shoe box
24. Open your eyes to make sure that the fit is perfect. If it is then you’re ready to take your shoes off
25. Now, without using your hands and keeping the shoes in the shoebox at all times, put both shoes back on your feet and tie the shoe laces (remember, no hands). Rummage around in your toolbox to find the right tool (the ABAOA tool should do the trick (Automatic Bipedal Accessory Overlay and Attachment tool). You can keep your eyes open now though
26. Remove the shoebox from your feet
27. Place the shoebox on the floor next to your right ear
28. You can now stop standing on your head
29. Stand inside the shoebox
30. Without bending your knees, undo the shoelaces of both shoes (any order is fine) using the ASD tool
31. Without using your hands, remove both shoes with the ABAD (Automatic Bipedal Accessory Disassociation) tool, making sure they remain in the shoebox
32. Step out of the shoebox
33. IMPORTANT! Make sure you apply all the changes so that they become permanent. If you don't you'll find the shoes are still on your feet and the shoebox disappears