Where is my Background? Let's use Gradient Descent to find it!
HTML-код
- Опубликовано: 21 окт 2024
- New collaboration with Gerrit Erdt to start developing a script to automatically define a region of interest that can be used as the background in processes like BackgroundNeutralization, ColorCalibration, SPCC, and continuum subtraction. The script is still in the beta version and Gerrit is working on further optimizations, scoring, and filtering in the algorithm.
I wanted to get the beta out to my viewers to try out, see how you like it, and what Gerrit and I can do to improve before an official release!
Get the standalone beta on my website: www.setiastro....
Looking for astronomy gear?
Use my links to Agena Astro or High Point Scientific and support the channel!
agenaastro.com...
www.highpoints...
Fantastic. You're developing incredibly useful tools add users in the Astro imaging community are taking notice for sure!
That's great feedback to hear!
Great script, and I like the idea how you circumvent local minima, the great downside of gradient descent.
Just wait until the full version! Gerrit is working on really making the scoring weights very robust. This was more of a proof of concept beta.
This is what I was really looking for. You are awesome! You rock!
Frank, love the image where the balls roll from the red to the blue. For someone like me who had to go to summer school for algebra, you have a way of making complex things seem simple. Bravo!
All your scripts so far are awesome................Bravo
Thanks so much! Trying to make things that solve problems I have and hoping others have similar issues to resolve :)
What a contribution again. You make our processes better and easier
You’re killing it man, thanks!
Many THANKS Frank !!!
Complimenti!!! 👏👍👏
Excellent! Good clear explanation too.
Thanks, this will save my eyes on the hunt
I don't know how I ventured onto this channel, but I'm glad I did. I really like the mathematical approach you use for Pixinsight. That is one thing I think has been lost in making the UI more friendly. When I started this journey on version 1.4, it was more command line based. Ultimately, the photos we capture are just data and I think it is important to know what you are doing to the data to understand how that affects the end result. It seems like a lot of people throw processes at an image and don't really know how or why it does what it does. I think this new tool will become part of my regular workflow as selecting the background is usually hit or miss for me, especially with unfamiliar targets. Thanks for the work you do to make the rest of us better! Do you use "buy me a coffee" or anything like that?
I am glad you stumbled upon my channel!! I am a very science and math orientated individual and I thought I can really share my knowledge and experience with the community. I think the more you know about the tools you have the better you can apply them!
You can see my website www.setiastro.com for contact and donation information.
This is very impressive. As a college student who just studied gradient descent in my multivariable calculus class this really hit home.
Can you explain more on how dark nebulosity is avoided? Usually it is darker than the background or at par with it. Also, what have you set as the termination condition?
Dark nebula isnt darker than true background, so no need to avoid it. The background region for colorcalibration just equalizes the RGB channels blackpoints to remove sky condition and pedastal differences between filters/sensor response.
For termination in this case is simple, if the neighboring regions average brightness is higher stop(there is also a check if there is a tie to use the region with the lower StDev) It is true there may be pathological examples where there is a loop of perfectly uniform median pixel vales and stdev it would loop around it forever, but real astro images arent going to have that.
Well you can tell your class there are real world applications for this! Lol
Excellent script!
You are a genius!
Thank you!
Great work. @Pixinsight must collaborate with you.
I wish! Lol.
Bravo! Thank you.
That’s cool! Since the GD is finding local minima(s) there is a chance it can settle on a dust mote (some still don’t take flats) or a stacking artifact or even dead pixel that wasn’t calibrated correctly/clipped.
It cant settle in a dead pixel. It is looking at a 50x50 region and scoring that entire region. There is a chance if you dont crop out your edge stacking artificats it can settle on that, so it is important you crop those out. As for dust motes, I cant help that... take flats :)
Good work!
yet another great stuff !!
Very nice!
Brilliant!
This is very handy! Saving yet more time for everyone.
2 questions:
1. Does Dark Nebula confuse/return false positives?
2. Now you have this automated, is it possible now to make the whole end to end continuum subtraction an all in one tool, potentially even a one click (or few click) experience?
Dark nebula still has background sky values so it wont confuse this method. For continuum subtraction Jurgens toolbox is already pretty close to a 1 button tool, but the color calibration method is a more rigorous method that now can be turned into a single tool. Sounds like a script for me to tackle :)
Been working on this if you want to try it out! I haven't released it yet, you can be the first! www.setiastro.com/s/ContinuumSubtractionv10.zip
One small tweak if you want the newest version
www.setiastro.com/s/ContinuumSubtractionv10-7xpw.zip
Will you let us know when it’s available in the repository?
Oh absolutely. That will happen when we finalize the script and you can bet I will make a video covering it :)
Newbee here. I have installed scripts in PI where I input a link into the repository but these are not like that. 'JavaScripts?". How to I install in PI?.
All my scripts are available via the repository. Even the updated FindBackground script Go to my website, it has the URL there. www.setiastro.com/pjsr-scripts