Go Back   IceInSpace > Equipment > Software and Computers

Reply
 
Thread Tools Rate Thread
  #1  
Old 18-02-2019, 12:54 PM
Stonius's Avatar
Stonius (Markus)
Registered User

Stonius is offline
 
Join Date: Mar 2015
Location: Melbourne
Posts: 1,495
Getting decent stars over nebulosity in PI

Hey all,

I'm finding good results in the nebulosity processing tends to crunch the colour out of the stars. Ultimately, I'd like them to be smaller, smoother and more colourful.

I suspect there's two parts to this.

When capturing I understand that the centres of the brighter stars will blow out if you want to capture the nebulosity.

Do people do a separate run with shorter exposures for stars?

Or if it's all just careful post processing, what is the best post process in pixinsight?

Is it just 'star-mask' or 'masked stretch' (both?). De-ringing? I'm playing with tutorials, but nothing seems to be making much difference.

I'm finding this a bit overwhelming with a less defined path to follow. Just looking for some direction in terms of processes other people use that works for them.

Best

Markus
Reply With Quote
  #2  
Old 19-02-2019, 09:32 AM
sil's Avatar
sil (Steve)
Not even a speck of dust

sil is offline
 
Join Date: Jun 2012
Location: Canberra
Posts: 1,474
Quote:
Originally Posted by Stonius View Post
Or if it's all just careful post processing, what is the best post process in pixinsight?

Is it just 'star-mask' or 'masked stretch' (both?). De-ringing? I'm playing with tutorials, but nothing seems to be making much difference.
Yes to all that. you can separate the two and work on them separately and yes deringing it vital and painful. the processing always depends on the image itself its NOT a set of values and buttons to push. Its a series of steps that varies depending on noise, faint structures and stars in the image. Understand what the purpose of each step is, some make no apparent change to the image but influence a later step if they were done several steps earlier. Take notes and follow/repeat them, learning and refining as you go, its hard work. Last Thursday I took a set of shots over an hour and I've only just got my subs registered yesterday, with another preprocessing step to go before i can attempt an integration. THEN will come actual stretching processing and no nebulosity in this image. Time and patience. Check out the Light Vortex Tutorials if you havent already, essential reading and learning resource.

Edit: go less agressive in your steps. Often its better to run one process with subtle settings a few times than running it once with heavy settings. Make use of previews to test settings and pay attention to the introduced artifacts, they will always be there but maybe not visible but can influence or multiply with later steps. Remember, signal and noise both exist at different scales in each image, so dont try to overprocess in one go to tackle all scales at once, you can go easy with smaller stepps at each scale separately.

Last edited by sil; 19-02-2019 at 09:44 AM. Reason: added tips
Reply With Quote
  #3  
Old 19-02-2019, 02:12 PM
Stonius's Avatar
Stonius (Markus)
Registered User

Stonius is offline
 
Join Date: Mar 2015
Location: Melbourne
Posts: 1,495
Thanks Steve,

You're right. That's how I've been proceeding thus far, but I assumed maybe there was a quicker way, than bumbling along.

Best

Markus
Reply With Quote
  #4  
Old 21-02-2019, 11:07 AM
Stonius's Avatar
Stonius (Markus)
Registered User

Stonius is offline
 
Join Date: Mar 2015
Location: Melbourne
Posts: 1,495
Steve, I know this is a huge subject, but I wonder if I might impose upon you to illuminate me on which processes *look like they don't do much, but can come back to bite you in the 'Asterope' if you know what I mean - I'd like to know the ones to keep an eye on.

By way of example, I've found cosmetic correction rarely does much to my images - it either clips a lot of pixels or none, so I'm starting to leave that one more or less alone unless I see that it's doing something. Maybe my calibrations are taking care of that for me - what do I know?

Many thanks

Markus
Reply With Quote
  #5  
Old 22-02-2019, 11:16 AM
sil's Avatar
sil (Steve)
Not even a speck of dust

sil is offline
 
Join Date: Jun 2012
Location: Canberra
Posts: 1,474
I feel my cosmetic correction steps need a bit more work to avoid any clipping. specifically i was thinking of multiscaletransformation, atrouswavelet, etc.

Problem is in something like photoshop you usually work with an 8bit per pixel image, sometimes 16bit but the tools like levels and histogram only work with 8bit precision so changing a value by one in practice is changing something in image by 1/256th which is noticable to the eye since our monitor usually only show us 8bits as well. But with a stacked image in something like pixinsight you can have 32 or 64bits per pixel and changing a value by one makes no visible difference to us but the change is there regardless, meaning the (colour/lightness) distance between adjacent pixels has widened slightly . now a later step may stretch values (such as bringing out nebulosity) by say ten times, so now that tiny insignificant difference from earlier is exaggerated ten times. another processing step may alter this further, and so on. so by the end two adjacent pixels which should be identical or imperceptably different alone a gradient are now so far that when you make a jpg to post they are now very different and noticable. across an inmage this might mean you have banding where there should be smooth gradients and artifacts around high contrast elements like planet edges or stars.

Hope that makes sense what I'm getting at there.

In reality though image processing is a compromise. We cant eradicate noise, there is always signal (we want) and noise (we dont want). We can suppress the noise so we can stretch and manipulate the signal more than usual knowing the noise will return but we choose arbitrarily where to stop so there's an acceptable amount of noise and we're happy with the signal.

With my workflow its handy to note not only what the step does and what i expect to see from it (if anything) but also what negatives it creates (adds new noise, creates rings around stars, etc) so at a later step that creates the same negative effect I need to be mindful it can compound with the earlier one. It may cause me to evaluate the two steps to see if they are similar and if i should go heavy on one and light on the other, or if i can dispose of one of them completely and still achieve the positive changes I desired. To note the negative effects of a step I do a bunch of tests where i run adjustments at intervals up to extreme values so i can see what negative effect manifest and if at a linear or geometric rate. Sometimes I run my workflow a few times fully making just a single value hange when tweaking to find optimal values I find acceptable in the final image.

masks are a source of negatives themselves. we can't see exactly where the glow around a star ends and becomes empty background but when masking we need to have them feather out at the edges. this feather region is where disparity gets created when we process, its often noticable around the trapezium in the core of orion when someone masks to stretch the core down for detail.

isolating stars/galaxies/nebulosity/background with masks presents problems of isolating ONLY what you want so your processing only works on them and nothing else. In practice its often impossible, so again a compromise point is where you aim for which is unique to each image and not definable as a setting value. at least that I'm aware of.

From Gerard's amazing pixinsight videos I've learnt much about how PI works and that certain processing tools are predefined duplicates of each other and that everything should be doable via PixelMath expressions. Ultimtely I'd like to turn as much of my workflow into pixelmaths expressions and I am certain it can be done for most of it and work equally well regardless the image, with only a few "hands on" steps for specific masks and value analyses.

really whatever it takes to get confident in my workflow, especially with the poor source data i can take I may decide at the camera to change camera settings to sacrifice star shape for a little extra faint signal and be confident it'll process ok.

your workflow is up to you with what effort you are willing to put in. I do suggest people go through tutorials even if for other software to see what steps people take and hopefuly why they do them, then you can take ideas and add them to your workflow. The workflow should be pretty much linear in that you go through all the steps in sequence, not skipping any. But you may have different workflows for different targets. for example mostly i capture with dslr + lens so its widefield still images and regardless of whats in view i process the same. for planets, moon, sun its a scope and zwo camera capturing video and processing is different but still the same principle of creating a working linear series of steps that you can rely on to produce an acceptable end image. as new features and software become available and you learn new things the workflow will change, expanding and contracting as needed but never forget only adjust if it benefits the end result. The steps dont matter much, just how they improve your end result, so you're doing it correctly while your images improve I'm no expert just a perfectionist learner. I don't think many of my images are very good, but i am happy with what I was able to achieve with poor initial data and satisified it the best i could do at the time. what others think of my images is irrelevant, many others get far better shots of the same targets but I'm yet to see any of those using comparable gear.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 04:17 AM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement