Hi Folks,
My website, Cosgrove's Cosmos, has just published some new posts under its Tips & Techniques Section entitled: "Using Deconvolution In Pixinsight".
https://cosgrovescosmos.com/tips-n-techniques/using-deconvolutioninpixinsight-part1-introduction
This started out to be a simple posting, but it grew so large that I ended up breaking it up into a series of 7 installments to make it easier to parse:
• Part 1 - An Introduction
• Part 2 - An Overview of PFS and Deconvolution
• Part 3 - Workflow Considerations
• Part 4 - Preparing for Deconvolution
• Part 5 - Example - The Draco Triplet
• Part 6 - Example - Messier 63
• Part 7 - Example - Messier 31
Deconvolution in Pixinsight is a very powerful tool for restoring lost detail in an optical system - but it is not the easiest tool to learn to use. When I started, I had a hard time coming up to speed - a problem that is all too common. But with time, I finally learned to use the tool to improve my images.
In this seven-post series, I share seems to work for me, and covers the background of the problem domain so you can understand not just the WHATs and the HOWs - but also the WHYs as well.
Covered are the concepts of Airy Disks and Point Spread Functions, considerations when estimating the PSF model for an image, Workflow considerations, how to create the support images needed by Deconvolution, and how to iterative testing to set needed parameters to get the best results. Finally, I go through three real-world examples using my own images.
I hope you find this helpful. I will be evolving this series in place as I learn more and as I get feedback and suggestions for improvement. ( I will be adding a change log to Part 1 to cover updates).
Thanks for looking!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Good work Patrick!
Can't wait to read this information.
Thanks for sharing this buddy.
Mike
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
|
Thanks for sharing Pat! I'm now subscribed to your website.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Patrick,
This is incredibly helpful and very useful.
I alway use deconvolution for my images at 640mm focal length, and still find that I get an improvement. I pretty much follow all your steps, including starting at pretty much similar values in the Starmask and Deconvolution routines. I have also experimented with RangeSelection + Convolution + CloneStamp to produce the LDSI. Do you have any thoughts on that approach?
I look forward to seeing your note on the EZ Deconvolution script. I tried it out, but never got quite as good results from using the individual PI routines.
Once again, well done for providing such a comprehensive, engaging and easy to follow primer on deconvolution.
Brian
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
|
Exactly what I need! I am learning PI at the moment and was contemplating attempting deconv. Thank you
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Tom Boyd:
Nicely done Sir !-)
Hi Tom! Thanks! Nice to find you on Astrobin! I just followed you!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Elmiko:
Good work Patrick!
Can't wait to read this information.
Thanks for sharing this buddy.
Mike
Thanks, Mike! Hope you find it helpful!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Jeff Kisslinger:
Thanks for sharing Pat! I'm now subscribed to your website.
Thanks, Jeff! Glad you have you on board!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Brian Boyle:
Patrick,
This is incredibly helpful and very useful.
I alway use deconvolution for my images at 640mm focal length, and still find that I get an improvement. I pretty much follow all your steps, including starting at pretty much similar values in the Starmask and Deconvolution routines. I have also experimented with RangeSelection + Convolution + CloneStamp to produce the LDSI. Do you have any thoughts on that approach?
I look forward to seeing your note on the EZ Deconvolution script. I tried it out, but never got quite as good results from using the individual PI routines.
Once again, well done for providing such a comprehensive, engaging and easy to follow primer on deconvolution.
Brian
Thank you so much Brian!
LDSI is really just a map covering the brightest stars - the ones that are likely to be saturated - in your image. You certainly could do the rangeselection approach and get what you want since the stars you are trying to get at the top of your range and most nebulae would be well below those levels. The only problem I see with that approach is that it is driven by intensity and not shape. Running Starmask, with all parameters being the default except for Levels being set one or two higher, does a pretty good job for this and is more sensitive to shape. It's simple to do and the only issue I ever had with it is that it will occasionally miss a star buried in a nebula or a galaxy body that really needs to be on the map - and when I see one like that, I just add it ib by hand. This does not happen very often though.
I am currently working on the EZ-Deonvolution article now. I have never been a user of the tool before this, but I must say that I have been very impressed with it for several reasons. I hope to have the next article done in a few days.
Thanks again,
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi
Stuart Taylor:
Exactly what I need! I am learning PI at the moment and was contemplating attempting deconv. Thank you
Hi Stuart,
I hope this proves to be helpful to you!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi Patrick,
this is very helpful indeed, would love to have had it 2 years ago, would have saved me lots of time! Regarding the LDSI, I often use the Maskgen script by Herbert Walter in the process, often in combination with advanced star mask. Maskgen creates a star mask based on the Gaia EDR3 catalog, so point like sources which we know are stars. that helps to include bright stars without the need of clone stamping, and also works very well for low contrast situations like stars on top of bright nebulous structures (where dark ringing is particularly visible). It also helps avoiding to get the center of galaxies misidentified as stars (something which starmask, starxterminator and starnet often do)
Matthias
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Thank you for this Pat, I’m on the page now and really look forward to your update “postscript” on the EZ-Decon process. Thats what I have been using to integrate this process in my workflow.
Dale
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Dale Penkala:
Thank you for this Pat, I’m on the page now and really look forward to your update “postscript” on the EZ-Decon process. Thats what I have been using to integrate this process in my workflow.
Dale
Hi Dale,
Hope this was helpful. The post on EZ-Decon will be released probably later today.
Clear Skies,
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Dale Penkala:
Thank you for this Pat, I’m on the page now and really look forward to your update “postscript” on the EZ-Decon process. Thats what I have been using to integrate this process in my workflow.
Dale
Hi Dale,
Hope this was helpful. The post on EZ-Decon will be released probably later today.
Clear Skies,
Pat
Great, I look forward to it!
Dale
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi Patrick,
this is very helpful indeed, would love to have had it 2 years ago, would have saved me lots of time! Regarding the LDSI, I often use the Maskgen script by Herbert Walter in the process, often in combination with advanced star mask. Maskgen creates a star mask based on the Gaia EDR3 catalog, so point like sources which we know are stars. that helps to include bright stars without the need of clone stamping, and also works very well for low contrast situations like stars on top of bright nebulous structures (where dark ringing is particularly visible). It also helps avoiding to get the center of galaxies misidentified as stars (something which starmask, starxterminator and starnet often do)
Matthias
Matthias,
Thanks for your note - and thanks for your kind comments. I have not tried that script but it sounds like a great way to go - I will be sure to check that out!
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
This is really a fantastic introduction to deconvolution in PixInsight, much appreciated.
Thank you and clear skies
Wolfgang
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi Pat,
I want to sincerely thank you for this series. You put in a lot of hard work putting this presentation together and I am very grateful!
I have been a PI user for just over a year and I have been spooked by Deconvolution. I have been using the EZ script with varying success on my images. Sometimes it works well and sometimes I just cannot get things right and give up.
I started with your Deconvolution series early this morning and have been working with it throughout the day. I just finally got a good decon process on my luminance image.
Very pleasing results and I really like how the stars are sharpened and reduced!
Now, I need to figure out how to integrate the deconvoluted luminance image back into my osc rgb image. You mentioned "Use LRGBCombination to fold the Synthetic or Ha “L” image into the color Image". Maybe separate the original osc image into R, G, and B using Channel Extraction and then combining the deconed L back in using LRGB combination???
Another question, you mentioned to do linear Noise Reduction to “knock the fizz off". Where is that process? It it MLT?
I also just subscribed to your web site. Thank you very much Pat!
Jim
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Thanks for posting this great resource.
Ian
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
This is really a fantastic introduction to deconvolution in PixInsight, much appreciated.
Thank you and clear skies
Wolfgang
Thanks so much, Wolfgang! I am happy you found it a help!
Clear Skies,
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi Pat,
I want to sincerely thank you for this series. You put in a lot of hard work putting this presentation together and I am very grateful!
I have been a PI user for just over a year and I have been spooked by Deconvolution. I have been using the EZ script with varying success on my images. Sometimes it works well and sometimes I just cannot get things right and give up.
I started with your Deconvolution series early this morning and have been working with it throughout the day. I just finally got a good decon process on my luminance image.
Very pleasing results and I really like how the stars are sharpened and reduced!
Now, I need to figure out how to integrate the deconvoluted luminance image back into my osc rgb image. You mentioned "Use LRGBCombination to fold the Synthetic or Ha “L” image into the color Image". Maybe separate the original osc image into R, G, and B using Channel Extraction and then combining the deconed L back in using LRGB combination???
Another question, you mentioned to do linear Noise Reduction to “knock the fizz off". Where is that process? It it MLT?
I also just subscribed to your web site. Thank you very much Pat!
Jim
Hi Jim,
Thank you so much for your note and your feedback. I am really glad the posting helped you make progress with deconvolution.
If you have an OSC image, you already have the color image. I would do the linear processing (which for me is DBE, PCC, and a linear denoise. Then I stretch the image to make it nonlinear. Once you have processed the L image, I would take this nonlinear. At that point, you can use LRGB with the L image listed in the L field - and leave the RGB fields empty. drag the LRGBCombination triangle to your color image and you are all set!
As far as linear denoise, I sometimes use MLT, sometimes I use EZ-Denoise, and sometimes I use Demure Denoise. I sometimes try all three and see which looks better to me. Look at the detailed processing notes on my last 10 or so images and you can see what I did in each case.
Hope that helps!
Clear Skies,
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Ian Dixon:
Thanks for posting this great resource.
Ian
Thanks so much, Ian!
CS,
Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Hi Pat,
I want to sincerely thank you for this series. You put in a lot of hard work putting this presentation together and I am very grateful!
I have been a PI user for just over a year and I have been spooked by Deconvolution. I have been using the EZ script with varying success on my images. Sometimes it works well and sometimes I just cannot get things right and give up.
I started with your Deconvolution series early this morning and have been working with it throughout the day. I just finally got a good decon process on my luminance image.
Very pleasing results and I really like how the stars are sharpened and reduced!
Now, I need to figure out how to integrate the deconvoluted luminance image back into my osc rgb image. You mentioned "Use LRGBCombination to fold the Synthetic or Ha “L” image into the color Image". Maybe separate the original osc image into R, G, and B using Channel Extraction and then combining the deconed L back in using LRGB combination???
Another question, you mentioned to do linear Noise Reduction to “knock the fizz off". Where is that process? It it MLT?
I also just subscribed to your web site. Thank you very much Pat!
Jim
Hi Jim,
Thank you so much for your note and your feedback. I am really glad the posting helped you make progress with deconvolution.
If you have an OSC image, you already have the color image. I would do the linear processing (which for me is DBE, PCC, and a linear denoise. Then I stretch the image to make it nonlinear. Once you have processed the L image, I would take this nonlinear. At that point, you can use LRGB with the L image listed in the L field - and leave the RGB fields empty. drag the LRGBCombination triangle to your color image and you are all set!
As far as linear denoise, I sometimes use MLT, sometimes I use EZ-Denoise, and sometimes I use Demure Denoise. I sometimes try all three and see which looks better to me. Look at the detailed processing notes on my last 10 or so images and you can see what I did in each case.
Hope that helps!
Clear Skies,
Pat
It absolutely does help Pat. Can’t wait to get back to the pc!
Thanks Pat
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Well done Pat! I went through the 7 installments, and I found them extremely interesting.
I wish I had such a high quality resource when I started using Deconvolution and EZ Decon!
Thank you for producing sharing such a great tutorial/reference.
CS, Mau
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
12.44
3.34
9.53
2.11
16.18
4.69
0.90
5.87
15.85
7.75
6.10
1.20
