BlurXTerminator with Russ Croman (Highlights/Shorter version)

Free 30-day trial (or purchase) of BlurXTerminator:
www.rc-astro.com/resources/Bl...
THE FULL INTERVIEW:
• BlurXTerminator with R...
Here is the first interview Russ has given since he released BlurXTerminator to the world in late December 2022. This is a wide ranging interview that covers many areas and addresses many questions through Russ's explanations and demonstrations.
As mentioned in the video a comprehensive explanation of the architecture of Russ's neural network can be found in this AIC video he published on his own KZbin channel at: • AI for Astrophotograph...
Use the Chapters below to find topics of interest:
00:00 Introduction (Interview and Explanation)
09:00 Tour of the BXT process by Russ
10:08 Explanation of Automatic/Manual PSF for NonStellar features
12:15 Over sharpening (Fine scale deconvolution)
13:24 Determine Manual PSF Diameter
14:49 Improvements and AI version 2
16:39 Corrections of aberrations (demonstration)
17:56 Correct Only
19:42 Luminance Only Mode
21:19 WorkFlow of Processes
26:52 Baffled at BXT Popularity?? (SNR example)
31:00 Ill matched PSFs for LRGB (example)
32:26 SH2-104 example (artifacts are small))
36:47 BXT only sharpens features that exist in image
39:07 BXT does not hallucinate structures (not content aware)
43:14 Results (vdB 4)
45:42 Wrap Up
More at:
AdamBlockStudios.com
See the Reviews:
www.adamblockstudios.com/cate...
Those just starting out in PixInsight will enjoy FastTrack Training!
(Updated to include SPCC!!!)
www.adamblockstudios.com/cate...
(see the video • Introducing FastTrack ... )
#PixInsight #BlurXTerminator #NoiseXTerminator #StarXTerminator #Deconvolution #neuralnetworks #vdb1 #vdb4 #sh2106 #AdamBlockStudios

[@lukomatico]

What a great interview Adam! - Huge thanks to you and Russell for taking the time to put this together, - also, very glad you asked about future projects haha, the tentatively titled 'GradientXTerminator AI' sounds like it'll be the next big thing! 🙂 Great work, much respect to you both!

[@ianspector1357]

Russ Corman's products have such integrity, they really have been developed to get the best from our data. I bought BlurXterminator just two minutes into the original launch video, I chose an image and just dropped it onto it with default settings. I was stunned at how good it was. So much so I have been going back and reprocessing last year's data. I never go back, but this tool is so profoundly wonderful it is changing my approach to star reduction as well as deconvolution. All of Russ's tools are a joy to use but this one is simply remarkable. One suggested improvement, if you charged more then I would not feel so guilty using it so much.

[@mcastrophotography76]

Thank u Russ, u're a genius ! A hug to all those who "pix deconvolution is better".

[@scidav87]

I'll watch when I get home. Great subject and explanation. Thank you!

[@slmdaddy1]

Thank you Adam - lots of excellent information. Appreciate it.

[@AdamBlock]

Thanks!

[@jeffreyweiss5615]

Great conversational video that helps tremendously in understanding this terrific new tool and the subtleties of its use. Thanks Russ and Adam.

[@Aerostar509]

Thanks for showing the best order to use all these processing tools.

[@filipgloria]

Great interview! This is such a revolutionary thing you’ve developed Russel! Maybe very specific, question for Russel: When I take three monochrome images, I usually have the problem that the errors in seeing/guiding/aberration can be slightly different from each other. If I have some elongated stars only on the oxygen channel, when I use BlurXterminator to correct the images, the stars sometimes have a tendency to be corrected in such a way that the AI “thinks” that some of the stars are in a slightly different place than on the other channels. The result is that you get almost like an “aberration’-esque result. I’m not sure if thats what happening, but I’ve just noticed that the corrections can sometimes be a little different on each channel, resulting in some of the stars looking “wonky” when putting them together.

[@rrcroman]

Have you tried channel combination followed by BXT? That's the best way to ensure that the neural net can see all channels at the same time and make any aberration corrections.

[@indysbike3014]

I can open a non-stretched image in Photoshop so please make it available for Photoshop (or Siril perhaps).

[@AdamBlock]

For the time being Russ has indicated this is strictly a PixInsight application.

[@indysbike3014]

@@AdamBlock I know, just hope it becomes available for Photoshop as not everyone uses Pixinsight. I'm doing fine with Siril, Photoshop and Russ Croman's plug-ins.

[@cosmoscarl4332]

What would be a great desktop computer, right out of the box, for image processing in Pixinsight?

[@julianmatthews3262]

Fabulous interview Adam and thanks to Russ for this process. You say that you need sufficient good data to really see the results of the blur run. I have tried on two different images one with 5 hr and the other 6 hrs of data with OSC stacked calibrated images from my 294c. The stars look fab but I do not see much difference with the nebula images. Am I too short with my integration time to see good results to the nebula and if so whhat should I be aiming for? Thank you

[@AdamBlock]

You didn't say what the object was..but I can say for a given set of equipment under a given sky... not all things are possible. There are limits. It could be no integration time will meet the bar. Sorry for the bad news. Darker skies are usually the answer... then bigger telescopes... better cameras.. .LOL.

[@julianmatthews3262]

Apologies Adam, the California Nebula and the Jellyfish Nebula. My skies are Bortle 5 and I was using a quadband filter from Altairastro through a 102mm refractor.

[@AdamBlock]

@@julianmatthews3262 Yeah... all of that sounds like a challenged situation. Bright skies, small aperture and effectively narrowband imagery. I would not expect these nebula to have enough signal to benefit directly from BXT with your situation. Some large telescope data and dark skies would help.

[@timothyrashmawy4251]

Thank you for the video. It’s very informative. So, like the guy you referenced who said he was baffled at the praise blurxterminator has been getting, the image I have been working with has the same noise issue. It sounds like from what you were saying in the video, there’s really nothing I can do to be able to use blurxterminator for sharpening of this particular image. Am I understanding this correctly?

[@AdamBlock]

Correct. You have to be tall enough to be admitted to the ride my friend. You need more signal... and for a given set up of equipment and location it simply may not be possible to achieve. Most things you can spend more time to get enough exposure time and minimize the noise to get thinks to work. Also smooth/fuzzy things will remain smooth/fuzzy (as they should).

[@ianspector1357]

Have you tried his noiseXterminator? It is quite remarkable what it can do to noisy images.

[@laurentandre6139]

@@ianspector1357 I was wondering the same thing... maybe (just maybe) in those "high noise cases", using NXT before BXT could be usefull ?! => to put to the test

[@laurentandre6139]

WorkFlow wise, What about applying DBE > BXT > NXT to all R/G/B/L images Before color combination ? (i saw that spcc seems worst off after BXT... but... overall ?) To smoothen color noise & bring color structures up as well ?! PS: Hi all & thanks for this video & for those great RC products (After tryout... i just licenced them all !)

[@AdamBlock]

This would use mor time (since many of these processes operate on all channels of a color image in the same way as if you did them individually) and it would negate a benefit of BXT. BXT looks a the relationship in color between channels to correct for aberrations. You would not have this benefit if you worked on the channels individually. Finally... if you did not make absolutely certain BXT was the same on each channel... you will have problems.

[@laurentandre6139]

@@AdamBlock Thank you for your answer Ok, got It on the RVB "synergy" plus. In a complex L + R/V/B + H/O scenario, where we usually go the StarNet (or SXT) route on L quite early in the process, i'd go on BXT before star removal for BXT's algo sake, on the side, plus on each of the H&O chanels to maximize algo awerness on these narrowband signals before adding then to the color (with less S/N) mix. ~as a compromise ?

[@leomarchadier1239]

BlurXterminator is more efficient on starless for sharpen nonstellar