AutoDBE: Throwing my hat in the gradient correction ring, for good reason!
Vložit
- čas přidán 9. 06. 2024
- Join this channel to get access to perks:
/ @setiv2
I think most people know that other auto gradient correction tools can try to correct small scale structure that isn't gradient, but maybe you don't know how prevalent that actually is! I also feel DBE if you spend the time meticulously placing points can get you the superior result.
Wouldn't it be nice if there was a way to have those points created automatically? Maybe also have a means to exclude dark nebula so they don't become part of the gradient correction model??? That is exactly why I created this script!
Get the newest gradient correction tool ADBE and put it in your toolbelt to use!
Links to my website: www.setiastro.com/pjsr-scripts
Be sure to leave and comments and suggestions!
Big shout out to my channel members for the help troubleshooting and testing this newest script! - Věda a technologie
Awesome! I saw this script about 10 mins ago get added to my PI update and I was thinking "there has to be a video or something associated with this." Just got this video update 2 mins ago, so crisis averted.
In all seriousness, thank you. Auto-DBE is incredible. This is a great new tool for us amateurs. I love how you've identified these extracted backgrounds that are taking structures with them. You've given me ANOTHER reason to go back re-do all my images.
The comparison to Graxpert was shocking. Your scripts are killing it.
I have to try this on old data and compare! I love the fact that it’s not a blackbox and we can look ‘under the hood.’ Amazing work. Thank you so much.
Looks amazing! Will need to test it out :) In particular being able to check the DBE points is great. To be fair to GXP, while its creators recommend smoothing=0, from my tests I've recommended from the beginning to set it to 1 in most cases (unless there are sharp gradients such as Newt reflections in the image), it's far, far more respectful of structures at smoothing 1!
By the way I used your continuum extraction script as well as your statistical script on a live stream yesterday, everything worked amazing!
That awesome Cuiv! Didnt even know you knew about my channel! Long time fan of yours! Followed your tutorial to set up my MeLe!
You are doing real great job to help astrophotographers like us. Highly appreciate you for making our life easier. A big thumbs up.
Thanks Franklin you are winning at script writing currently! Will definitely be adding this! 👏
After watching the video yesterday I spent the evening extracting backgrounds with Graxpert and your script to compare. Your script does an excellent job is my conclusion. Thanks for doing this, it is so helpful!
This is an awesome contribution to the Pixinsight DBE workflow!
Impressive again. I always struggle with gradient in my polluted zone. Thx
Totally agree re Graxpert stripping out structure. Ran it on Rho Ophiucus and the extracted background looked like .... Rho Ophiucus! Same with the LMC. I always liked the DBE in APT, but that is a slow and artistic process. So this inside Pi is a great step forward. A huge thanks for this.
You’re an excellent presenter….thank you for all that you do😊
Fantastic - I was kind of expecting this, it seemed a logical step from 'find my background', but the results look even better than I expected.
Great! It's always good to have alternatives. Thanks a lot for this interesting script and for your time developing those very helpful tools.
I was mostly using DBE cause with AI I wasn’t happy. Now yours is a super duper script! Will install it tonight! thank you a million!
Yes I have gone back to DBE after using GraxPert and Graident Correction, but I will now be using your new AutoDBE, thanks mate
That's amazing Franklin! I need to try the script out! I always had the impression that Graxpert is removing some information from nebulae and galaxies but I was unable to find out if my feeling is correct. Therefore I switched over to use the PI Gradient Correction tool more often. The Gradient Correction tool seems to remove less structure from what I could see on the screen. Any step towards affecting the gradient itself only is very much welcome when shooting from a strongly light polluted area.
PI gradientCorrection is even worse!! Here are two videos I made on that
czcams.com/video/tq-iXsElFUw/video.htmlsi=T-MbqYh3KxHeLiOM
czcams.com/video/VBBMHIOZQ8Q/video.htmlsi=DfTIMXTb3hVkyaBI
All the automatic gradient removal tools seem to remove structure, that is why I am hoping mine does a better job
Until 15mn back i was completely frustrated the way AI gradient correction tools are messing with my background and had no intention to learn DBE. Now i'm all over my old folders running this script and i must say this is AWESOME. Keep up the good work, Franklin !!
Absolutely fantastic work again Frank. The speed you’re going at you may end up building your own version of Pixinsight altogether. Can’t wait to see what’s next. In the meantime I’ve got some images to ADBE now 😉
Keep up the great work!
Blown away by all your scripts! They have become an essential part of my PI workflow! Thank you!
What other scripts do you recommend?
@@indysbike3014 I use Star Stretch and Statistical Stretch regularly... great helper scripts
Wow, superb and I have had my doubts about these latest gradient tools, and you have proved my theories….nice new tool..👍🏻
I would encourage anyone to take a hard look at the gradients either GradientCorrection or GraXpert pull out just like I did.
Even yet more brilliant work, saving us all time & saving data in our images too. Huge thankyou!
Thanks! We still need to get your What's In My Sky script or program rolling!
Another great script Frank. I find I at least run a DBE on most images to compare with Graxpert and more often than not use the DBE version (Does take time as you say). I use all your scripts and have no doubt I will be using this one as well
Wow! You're killing it with your scripts. I've always kinda just gone with ABE with function degree to 2.
Tried GraXpert a while ago, but the software left a "brain pattern" in some of my images and after I noticed that, I stopped using it entirely and went back to ABE.
Yours looks to be a massive improvement. Good job on these, and thanks for sharing them with the community. Much.. MUCH appreciated.
Thanks so much!
This is so nice, thanks for creating this tool, will add to PI tonight :)
Really nice script, I’m looking forward to giving it a try!
The results look really impressive! I'm looking forward to trying this out.
Fantastic script Frank !!!
Great, will try it tonight. Thank you!
Again, thanks very much for your amazing work.
Awesome Work ! Can't wait to try this out tonight !
Great Work! Seeing what you've done, and made available to the community, no brainer to join your channel to support the effort. Thank you!
I can't believe how easy this script is
This is a great help
So much better results than what I've been able to do
Thanks.
You provide us several awesome tools, and make my life easier. Thanks for your great contribution!
very, very good idea! Auto-DBE is an incredible script.
Your scripts are awesome!! This script is GREAT, i use Graxpert but after using your script it has made a whole lot of difference. You do amazing work , thanks for your in detail videos 💯
Absolutely crushing it!
It works great thanks frank!!
Looks VERY interesting!
this is a very nice tool mate, thank you very much (:
Thank you Frank for your enormous efforts in the interest of the astro community. Despite the good results with ADBE, I think a little criticism regarding 2 points is appropriate: 1) please avoid new updates every day, rather collect a little and release a new one a month max. 2) the developers at Graxpert have done a great job and are also making their tool available to this community free of charge. I think we should be a little more respectful in comparisons and give the other developers the chance to react to possible improvements in advance, e.g. by exchanging ideas without the community first. Just a small attempt for fair competition in an already crazy world. Thank you again, CS Andreas
Thanks for the feedback. I too make all my tools free of charge. I am also not the first to point out items that graxpert removes. That is well documented. As for the pace of updates, I try to fix issues as they arise. I get a lot of feedback for corrections and additions. I think a month would be a very long time to fix issues people are facing. I am travelling for the next week and a half so at least I wouldnt expect any update then :)
😂 have a nice travel and don‘t worry about my comments, you‘ve done really a great work 👍 CS Andreas
fantasic job - thanks for all ýour scripts
Finally someone put words to what I have been trying to say to some club members. But I didn’t know how to describe it. Thanks!!
Hopefully now you can see how I was able to take a harder look at the structure in what these tools pull out for their corrected background. Pretty easy to slap ABE in degree 1 or 2 onto the background they pull out just to really see what is going on
This is really great. I'm no expert processor but I picked up the same issue with Graxpert
Awesome!
Wow. Thank you
Adam Block once said that if the detail is so faint, then why bother trying to bring it out. I sort of agree with him. Having said that, thanks for throwing another option into the ring.
Again WOW and thank you.
Very cool!
Excellent
Very nice script! I’ve also thought about eventual benefits of doing an ADBE on the single subs before you stack them? Do you think that would be beneficiary somehow. The benefit would be that you could potentially remove gradients that dynamically change during the course of a night.
That is what Local Normalization does during WBPP. It would be time consuming, but you can calibrate your subs first, then run ADBE on all them, then use those as your subs to stack. Not sure how much better of a result you would get, if you like experimenting then this is right up your alley!
Wow wow wow... grazie!!! 💯👍
amazing thank you
Wow just run it on a a couple of files that I could not remove the gradient on so never processed the data…….. well until now. Thank you again excellent script
Thats great to hear!
Très fort 💪…!.
This is genius like all your scripts. Now I just need clear skies. Thank you again!
Nice work. This tool looks like it will be a fast and efficient way to solve some problems which only DBE or annoying masking could really handle well to this point.
Your Andromeda example is a pretty bad result. Much of the central structure it removed belongs to the galaxy and you end up quite over-corrected around Andromeda. I got a similar result with an Andromeda image I have, and it also ate up some structure on the side. But the very fast region rejection tool you included allows for a problem like this to be corrected quite efficiently, so it ends up producing a very good result with less work than DBE.
Definitely a new tool in my toolbox.
Awesome again. I have a question , though. What if we´d want to compare ,for a while, backgrounds from the script and from Graxpert?. Should I substract (with pixel math a simple - ) a background from the other?..If so....The background with less information would be the one that subtracted less data to my image?...Would you use another way of comparing both sets?... Thanks again for all your work. It`s amazing.
A simple subtraction will lead to clipping. Use GraXpertbacground - ADBEbackground + 0.1. Then run an stf on that result. Brighter areas in that result will be stronger subtraction from graxpert/less subtraction from ADBE. You should also run ABE on each background like I did in the video if you want to see what smaller scsle structure is being affected by both.
Wow that is impressive!!! I wonder could ChatGPT be used to GPU accelerate and existing script like WBPP which can be mind bogglingly slow on the biggest PC? Really love the scripts you have delivered - they are a big part of my go to now!
You will have to ask PI about fixing the WBPP speed with GPU acceleration :)
@@setiv2 I believe it is actually on their to do list somewhere - just way down under many other priorities and a small workforce - so don't hold your breathe!
Hello Seti Astro thanks for all your hard work,one question on dbe is it normal to have to click the 24 bit icon to be able to see the gradients?.
Be sure to run STF to see the gradients well. As for 24bit, i never turn my 24bit off. I think you cant see nearly as well. So i recommend in Edit -> Global Preferences -> Miscellaneous Image there is a checkbox near the bottom to default to always use 24bit and high dpi screen renderings
Looking forward to trying this out. I do have an issue though. Trying to download with your repository and Pixinsight (version 1.8.9-2) comes up with an error, skipping the first bit, "Parsing repository update information: Parsing Signature element (line=1 offset=0): Unkonwn code signing identity" followed by I assume your name. Tried removing/adding the repo, triple and quadruple checking the repo address, and resetting updates with no luck. I could just do the ones I want manually, but then I'd have the hassle of manually updating them too. Any ideas?
If you havent upgraded to PixInsight build 1605 (which came out in March) I highly recommend you do that as all scripts are going to require signatures going forward. You do this from the PixInsight download site. This will fix the error if you are trying to use the repository.
If you still run into a parsing error or signature error you can download the zip file directly from my website and then just extract the .js files (delete the .xsgn file). You can then go to Scripts -> Feature Scripts -> Add Script and select the folder you extracted the .js file too.
Let me know if this helps
Frank
This looks to be an awesome configurable tool. I am not a Pixinsight user though. Does this work as a stand alone or integrated in Siril?
It is strictly a pixinsight script. It automates various pixinsight processes to achieve the gradient correction.
Works really well. However it seems that the script in its current version strips a lot of the FITS header information. egain, binning, focal length, etc... (Edit: thanks for the amazing response time and "fix")
V1.1 out fixing this
Thanks for the useful script. The repository is not updating in PI. Somthing wrong?
Are you getting an error? What PI build do you have?
@@setiv2 Ver 1.8.9-2 no errors.
@@setiv2 Ver. 1.8.9-2 no errors. It shows as installed update, but nothing shows up in the Setiastro script box.
@@CMOS-Astrophoto what build are you on though? You need to be on the latest 1605.
You can also try going to scripts->feature scripts->add scripts. Then navigate to c://program files/pixinsight/src/scripts then click on Select Folder. PI will then look to add any scripts it may have missed during the update
@@setiv2 I have build 1605. Looks like the repository link did not add a folder for the A-DBE in the script folder. I added one manually and now it works. Thanks for the help :)
Is it possible to adapt this to be a plugin/script for siril? I'm not super knowledgeable on script making so I don't know of it is or not!
Probably not this. It uses core pixinsight functions to run. I havent used Siril before. If it has gradient correction tools that can be called via javascripts then it should be possible to do something similar, but currently this is strictly a PI tool.
@@setiv2 okay that makes sense. It does have a built in background extractor and I know siril allows for custom scripts but that's the extent of my knowledge. I understand if it wouldn't be able to be done but it would be awesome! Either way, great work man!
Downloaded Siril and looked at the code for their scripting. Unfortunately scripting in Siril is a very basic command call code. So it uses the commands line to just call the next process. Subkey is the command for the gradient removal tool and it would be coded like this: subsky { -rbf | degree } [-dither] [-samples=20] [-tolerance=1.0] [-smooth=0.5]. That in and of itself isnt bad, the problem comes from the loss of JavaScript functionality in this environment. In my ADBE script it is creating arrays in memory, reading sections of the image and doing lots of calculations in the background creating those arrays of points and then those points are applied to DBE. There isnt a way for me to do all that in Siril and then assign those points to Subsky for example. I will keep researching, but I don't think I can come anywhere close to what I'm doing in PixInsight right now in Siril.
@@setiv2 okay wow thanks you for researching! I'm glad to hear the background extraction isn't bad. It sounds significantly more complex then the very user friendly free Siril. If you are still researching though that does give hope you might feel it's possible. How hard would it be to create a standalone program like starnet? I don't want to put extra work on you but your program is amazing!
@@adamlong9031 i am not in software development and only just learned enough javascript to code for PI. I cant believe I am seen as a good script writer now lol. So dont hold out for me making a standalone program. Maybe I should reach out to the graxpert people and let them know how I am doing gradient descent for region selection. Maybe they want to include it in their RBF modelling. They seem all in on "AI" though at the moment
Can this tool run as a stand alone script outside pixinsight?
No, this is strictly a script within PI as it uses core PI functionality
I placed the link into Manage Repositories but it doesn't show up in PixInsight.
Did you get an error? Be sure you copy the full address including the final slash
@@setiv2 This the error code when I check for updates : Parsing repository update information: Parsing Signature element (line=1 offset=0): Unknown code signing identity 'FranklinMarek'
My Pixinsight is not able to download repository information. Getting an Error Unknown Code Signing Identity FranklinMarek
If you havent upgraded to PixInsight build 1605 (which came out in March) I highly recommend you do that as all scripts are going to require signatures going forward. You do this from the PixInsight download site. This will fix the error if you are trying to use the repository.
If you still run into a parsing error or signature error you can download the zip file directly from my website and then just extract the .js files (delete the .xsgn file). You can then go to Scripts -> Feature Scripts -> Add Script and select the folder you extracted the .js file too.
Let me know if this helps
Frank
Isn't a good deal of the structure destruction of these other processes due to running at low smoothness? Personally, I would never run with smoothness < 1, as most gradients are one variation or another of smooth shape (a linear wash, a circle radiating from a point, etc).
Exactly my thought. People are running Graxpert blindly without considering the settings, then blaming it for not working.
Graxpert works fine as a standalone tool for those that do not wish to pay the outrageous price of PI.
Hello I am a Beginner I don’t know, Where to download. This tool please.
You copy the URL address from my site, then go in PixInsight to Resources -> Manage Repositories then add. Paste the URL in the address field there. Then close that window. Go back to Resources -> Check for Updates. Install the updates it found. Then you need to exit pixinsight to finish the install
@@setiv2 thank you it is difficult for me to translate but I follow you now
I'll try your tool soon, looks very promising!
However, can you change or lower the volume of the music on your video, it's really annoying 😅 Always the same and very repetitive
50% of the people like it, 50% hate it. I will lower it more :)
I am not sure why, but I have the new version of PI, V. 1.8.9-3 build 1611 2024-06-19 and after I run ADBE using the defaults, Pixinsight freezes up on my laptop. My laptop is fast with Intel i7-12700-H, 64GB RAM, and a 4TB drive. Is anyone else having this issue?
I checked all my scripts right after the new PI update came out and all were working for me. When is it freezing on you? Right away after clicking execute or does it freeze after some time?
@@setiv2 I've just updated to windows 1.8.9-3 and statistical stretch will crash pixinsight as well as auto dbe, during execution for both or just even closing the script. It actually causes me to have to re-add scripts in feature scripts and make PI find installed modules such as BXT and NXT when I re-launch after a crash has occured. Haven't tried the others yet but it does seem rather odd
@@setiv2 It freezes after it is complete and the image is ready. Statisical stretch and star stretch work fine. I even closed everything else on my laptop.
@@setiv2 I ran it again twice. Once with "replace target image" checked and again unchecked. It runs and completes. It states COMPLETED, then PI says "not responding" in top left corner of PI screen. This is the only script so far that freezes on my laptop. Dell win 11, 12th Gen Intel core i7 12700-H 2.30GHz, 64 GB RAM, 4TB hard drive.
@@jims9982 in the console it says it is complete? Then it freezes? That is very strange! Can you run DBE itself?
Sorry but cannot get your scripts working ,either the main url address or downloading individual. In the individual there is only a Java script file a nd nothing else ,When adding script into script add on it doesnt find the script
If you havent upgraded to PixInsight build 1605 (which came out in March) I highly recommend you do that as all scripts are going to require signatures going forward. You do this from the PixInsight download site. This will fix the error if you are trying to use the repository.
If you still run into a parsing error or signature error you can download the zip file directly from my website and then just extract the .js files (delete the .xsgn file). You can then go to Scripts -> Feature Scripts -> Add Script and select the folder you extracted the .js file too.
Let me know if this helps
Frank
Bingo and thanks for the advice. ,all working fine . Maybe put something on the video stating this.
Forgive me for being critical but this is just a script for Pixinsight, not a "gradient removal tool" as it cannot be used without first purchasing PI at an eye watering 300 Euros.
Graxpert is free, standalone, and can be used with it's AI algorithm at various levels of aggression, OR can be set up entirely manually using three other algorithms, giving the user a wide range of options for best results.
Personally I find that I can run standalone Graxpert very well with almost zero removal of structure.
I agree this is a script within PI itself for those that already have PI. There are numerous PI only tools that are not standalone. Some free and some that cost additional money beyond the initial purchase of PI.
I'm a little confused about what you're doing here. Why do you believe placing DBE points at the 'blackest' places in the image is the right way to go?
My understanding of DBE is that you place the points on "background" (which should be all the same color/value) no matter what brightness is in the image and then DBE does some kind of fitting algorithm on all the pixels between the points you placed in order to calculate how to subtract/divide brightness/ADUs from your image to make all the placed points the same value (thus removing the gradients).
By only concentrating on the "darkest" regions of the image you're not telling DBE that the bright areas are also background.
What am I missing?
It doesnt go in the blackest per se. It finds the background regions across the whole image. To do that you necessarliy have to settle in the Local darkest region. Otherwise you will land on signal. Gradient descent finds local minima, not global. So once all those points are defined those are the background points.
Frank, the evolution of your gradient tools is masterful! Your contributions to astrophotography are game changers.