Can THIS finally DETHRONE BlurXterminator?? Free AI sharpening is AWESOME!

[@CuivTheLazyGeek]

My Patreon: www.patreon.com/cuivlazygeek My Merch Store: cuiv.myspreadshop.com/ Video about CosmicClarity: kzbin.info/www/bejne/lZybZJaGbsaGrLM GraXpert Release Candidate: github.com/Steffenhir/GraXpert/releases For any affiliate link purchases, make sure to accept cookies if prompted! And please be aware that extensions like Honey steal the referral! Amazon affiliate: amzn.to/49XTx01 Agena affiliate: bit.ly/3Om0hNG High Point Scientific affiliate: bit.ly/3lReu8R First Light Optics affiliate: tinyurl.com/yxd2jkr2 All-Star Telescope affiliate: bit.ly/3SCgVbV Astroshop eu Affiliate: tinyurl.com/2vafkax8

[@kleedhamhobby]

There's a video from Deep Space Astro on this where he suggests getting your FWHM value by using the Dynamic PSF window in Siril - Image Processing / Star Processing, click gear wheel to bring up Dynamic PSF window, click detect all stars (star symbol at bottom), click average PSF parameters (Sigma star), then last line in console will give FWHM value. I haven't tried any of this myself yet, but I intend to.

[@TonyUnderscore]

Just tried it myself and it works, many thanks!

[@MeIsAndy]

This is what I have been doing.

[@guyyanez6949]

That is correct. It is the way to do it using SIRIL as DeepSpaceAstro pointed out in his last video.

[@FocusAG]

Thank you very much for sharing that! Keep the good work :)

[@CuivTheLazyGeek]

Nice! Thanks for the tip!!

[@guyyanez6949]

Well done as usual! You are absolutely right, it does one deconvolution or the other and came up with the same conclusion that you provided. 1. Do the object only deconvolution, save result 2. Take the latter and do the stars only deconvolution.

[@CuivTheLazyGeek]

Thanks, good to know!

[@daflump6358]

Your channel has both cost me a fortune and also saved me a fortune !!! Thanks for letting me know about this new GraXpert feature, I love its AI denoise, very exciting to see such significant leaps !

[@CuivTheLazyGeek]

I hope it balances out in the end! Thanks mate!

[@coneman89]

Cuiv, I did it and pulled the plug on a Seestar S50 thanks to your videos! I used your affiliate link and just did my first photo of the sun seriously im impressed and look forward to tonight (come on clear skies !)

[@CuivTheLazyGeek]

Thanks mate!! Hope your first session under the stars went well!

[@coneman89]

@@CuivTheLazyGeek First session was amazing im blown away. Theres some features of the S50 that I haven't seen anyone mention namely Station mode I was already attracted to the telescope but things like station mode make it absolutely game changing for me

[@berndweisheit1157]

Don't forget: you compered a young Beta version with the finished produkt BlurXterminator!

[@CuivTheLazyGeek]

Absolutely!

[@jPaulofe]

Hi Cuiv! great video as always!! from the GraXpert 3.1.0rc2 download page we can see this info: "...Changes and improvements Added an ai model for stellar deconvolution. That is, deconvolving stars. The actual object (galaxy, nebulosity, ...) should not be affected by this model." So basically they seem to say that applying stars-only deconvolution should not affect the result from the previous processed image... Well...I'll try both and let you know while we wait for another (and better) rc or prod version 🙂

[@Fossbear]

crazy how Graxpert pushes forward. Even though i just bought BXT a month ago i still like the option of using different tools. Sometimes the standard workflow just does not give the wished result

[@CuivTheLazyGeek]

Congrats on the purchase though :)

[@mgscheue]

Having recently bought an S50, I'm impressed by the really good free software already available. Looks like some nice improvements coming soon (to Siril, too!).

[@CuivTheLazyGeek]

Indeed!

[@ignatiuk83]

That would be great to see comparison of free tools: graxpert vs Cosmic Clarity from SetiAstro

[@tomabxl4374]

I see a little dark hollow outline after the deconvulated stars with the graxpert… but maybe just my eyes. Probably can be fixed changing the values

[@CuivTheLazyGeek]

Yep, I see and remark on the same in the video

[@tomabxl4374]

@ yeah i saw at the end ( i noticed at the first picture I did not know you will also notice on the second) - great comparison, Cuiv, my friend!

[@ralfg9194]

This video and the sensible presentation of the software comparison is really valuable for us astrophotographers. Thank you very much for your never ending efforts, your competence and your willing and generous sharing of all this! Great work! Kind regards from Germany. If I may ask you for comments on these topics in one of your future videos: 1) Using AI tools (like BlurXTerminator...) with GPUs seems to need corresponding software drivers. Named are Cuda and TensorFlow. It is quite confusing if the Cuda libs have still to installed if one uses the recent installing process of the tensorflow libs or not. 2) If one has objects with low surface brightness that are on the edge of the surface brightness due to light pollution, are there any tips on how to separate them well in image processing?

[@donner5545]

Cuiv! Thanks for the comparison! So glad to know that the free software is starting to be in the ballpark of BXT! Just wanted ask about the pleiades image. Was there a field flattener used for that image? Seems like at the edge of the image that there's a color gradient in the stars (most apparent in the red and green layers), but I don't see the same effect as much near the center of the image. Do you think at all that BXT recognizes a potentially non-flat field and accounts for that, while the Graxpert tool does not?

[@CuivTheLazyGeek]

It was a poorly collimated Hyperstar image! BXT actually does local deconvolution (e.g. different PSF depending on the part of the image), which makes it so effective at fixing star shapes

[@jbreezy3146]

I believe that the two options for deconvolution is because some of us separate our images in two parts a starless and a star mask

[@CuivTheLazyGeek]

That makes sense but at the same time it doesn't make sense: how would the deconvolution algorithm estimate a PSF without stars in this image? Deconvolution relies on stars...

[@TheFredolegrand]

At 6:44 : no you are not missing anything : intermediate save is the workaround for this known bug of star vs object only deconvolution.

[@CuivTheLazyGeek]

Thanks!

[@deep_space_dave]

That's very good for beta! Competition is good! I personally have used DynamicPSF to get a good FWHM estimate when I was adjusting BlurXterminator manually especially when using it while stars were removed (yes you can do that but you need the PSF of the original stars). I believe BXT deconv the stars first then the objects by default. Maybe try that to see if you get a better result. Thanks for showing us this, I hope to see improvements that give a paid product a run for the money! Clear Skies!

[@CuivTheLazyGeek]

Thanks Dave! My issue is that if I do the star deconvolution, save the image, then do object deconvolution... The object deconvolution would run on a PSF that's not valid for that image (already deconvolved stars)! While BXT can just keep all of its local PSFs (since it splits the image in many sections) in memory!

[@rtpman1953]

Great test, Cuiv. Hopefully the developers are either aware of the issues or will incorporate your discoveries in the stable release.

[@CuivTheLazyGeek]

I hope so!

[@mbaztan_5814]

LOVE your videos dude!!

[@CuivTheLazyGeek]

I appreciate that!

[@fazergazer]

Interesting that Graxpert has an “object” deconvolution..normally I’d consider the stars as the local point spread function in a patch, and a resulting spatially-variable deconvolution should then handle each patch, stars and nebulousity…

[@michakubisz535]

Probably because AI was taught on the actual objects like nebulae and it recognizes them on your picture and measures the PSF

[@wadetubman]

new to astro with s50 following your video. This star sharpening has significantly improved my images. I also love graxpert denoise (as i seem to have a lot of noise). Just need to sort out stretching (not sure about graxpert stretches - are they for viewing only?)

[@CuivTheLazyGeek]

You can save the GraXpert image with the stretch if you wish! You also have the Seti Astro suite for stretching if you wish!

[@jmoreno6094]

The fact that BX reduces those nodules could be because it goes a bit further and after applies a soft blur (gaussian) filter. Just an idea. -You can try also to apply Graxpert twice with different fwhm - Graxpert doesnt have yet deconv for stars, thats why oblong stars are not well corrected Thanks for your videos Cuiv. Like!

[@JeffHorne]

Curious, which strength did you use for Sharpen Nonstellar in BlurX? Default? Great comparison!

[@cizirf143]

Hi Cuiv - same observation about stellar/object deconvolution. At this stage only 1 type of deconv is allowed per session... BTW, at this stage I prefer the star deconvolution of CosmicClarity as Graxpert tends to make stars slightly brighter and doesn't correct star roundness like BlurX does. On the other hand find GraXpert object deconvolution really good...

[@CuivTheLazyGeek]

Good to know! Thank you!

[@lucre73]

Thanks Cuiv!!! This help a lot. I'm trying non the new fatture with your suggestions (save the star decon image and re-opened in graXpert to do object decon) and i've checked in Siril if after decon the FWHM has changed....and it is! So i'm performing oblect decon with the "new". Hope it works as best as the star deconvolution

[@CuivTheLazyGeek]

In theory, the object decon should be done before stars decon, since it relies on star shapes to find the PSF to deconvolve with! And if the star shapes have been altered, the object decon will run on an incorrect PSF!

[@jacoblee679]

Looking forward to Graxpert intergration into Siril in the next version

[@CuivTheLazyGeek]

Yesss!!

[@caroliensche13]

17:50 it was stated that flux is invariant to the latest KI version of BXT, so the integral brightness of a star remains.

[@CuivTheLazyGeek]

Thanks for the precision!

[@magicmountainpercussion3336]

so glad to see this! great info! I just wish there was an image with elongated stars to see how well it actually does. my images last night using my gti mount and no guide scope had so elongated stars. using a dslr for now with a nikkor 70-200 is pitiful! I am going to try graxpert to see of o can improve my images enough to be usable. thanks again, you rock Cuiv!

[@CuivTheLazyGeek]

There is though? The Pleiades image at the end of the video, with the stars in the corners

[@magicmountainpercussion3336]

How did I miss that...durp! Thanks, I downloaded the windows version of graxpert and I need to find where it is on my computer...it socks to get old! I am working on a huge image of m 42 with 4 master files to create an hdr image after I remove the stars. Then I will use the shorter exposure image with the cleanest stars and use that image for the stars.

[@prasejonson]

If they just add star extraction, graxpert will be perfect! Awesome video as always

[@KevinRudd-w8s]

I will give it a try. Considering it is a free piece of software it seems to do a reasonable job. I'm still having mixed results with Cosmic Clarity, for some images it seems to work fine, for others the sars can end up worse, particularly at the edges though in the centre they usually do look better. Thanks for this update on Graxpert. Clear Skies.

[@CuivTheLazyGeek]

Always good to see progress in free software, thanks for the details on CC!

[@creazie]

thank you, exactly what I wanted to see !

[@CuivTheLazyGeek]

Cheers!

[@anthonylitchfield4853]

Can you recommend any additional filters useful for the Seestar that would produce different results for certain subjects

[@philleng480]

Great job Cuiv and thank you. Still think itbpays to test all of the gradient tools. Graxpert tends to rip out the nebulosity, but not always. Gct in pi can do a good job, sometimes. But wierdly if you dont mind a slow process then APP does a great job - although subjective, do you think there is nebulosity here or here? Thanks Cuiv.

[@CuivTheLazyGeek]

I almost never get any nebulosity ripped out at smoothing of 1.0 (I have videos on the topic), and I haven't been able to get the MGC in PI 1.9 to work well on most of my images...!

[@vaibhavsati538]

Astrophotography has always been pay to win in terms of equipment as well.

[@qwertymonsta]

pay to win is a funny term, but yeah... usually more expensive equipment gives better results. Otherwise, people would just go with whatever is cheapest lmao

[@vaibhavsati538]

@qwertymonsta yeah right and access to dark skies is pretty imp too

[@CuivTheLazyGeek]

True! I used the term in a facetious manner but it seems to have hit a nerve!

[@mbaztan_5814]

It looks amazing,i Will try graxpert and seems to be a quite good free option

[@Alan-vk6bk]

As I paid for BXT I'll be sticking with it....

[@CuivTheLazyGeek]

Sounds good to me!

[@richardwills6930]

As said above , I'm doing better with cosmic clarity, although as far as I'm aware, the developer of cosmic clarity has shared with graxpert, so I'm not sure why they're not on par . But I can't complain that people are willing to use their time to come within touching distance of blur exterminator. The future should be good and free !! 😂

[@CuivTheLazyGeek]

Frank keeps making his Seti Astro Suite better and better!

[@joshuamuller9705]

Have to try it. Thanks a lot 🙏 I like they give you option to use decon on stars or the object. It will improve over time. 👌 edit: did hear about the honey stuff from Luke? 😅

[@CuivTheLazyGeek]

Thanks mate! It definitely will improve! Oh and I told Luke about Honey in the first place, after the megalag video! lol

[@conradcardano1305]

Graxpert vs. Blur Extermination. Hmmmmm. When the images are unmagnified, they both look the same. I only see the difference on the highly magnified images. Does anybody ever print out highly magnified images? I don't think so. I think a big disadvantage with Blur Exterminator is that is requires Pixinsight. I will try the new version of Graxpert. Well done!

[@certhass]

I wouldnt say pay to win, i mean the image processing software also costs money. its a tool like your telescope and you pay for what you get. better telescope, better mount, better software... and you still need skill and practice to use all the tools :)

[@CuivTheLazyGeek]

Of course :) I was being a bit facetious!

[@MG-qj2zh]

Nice comparison, but my bias is towards SAS with CosmicClearity

[@CuivTheLazyGeek]

I compared it to BXT a couple months back, need to check it again!

[@MG-qj2zh]

@@CuivTheLazyGeek I'm looking forward to it kzbin.info/www/bejne/h2S6Zoqtlq55fJI SAS and Cosmic have come out in new versions. The star deconvolution in GraXpert is currently not really usable and the developers write that themselves.

[@jeroenfrijhoff8392]

Unfortunately so far I only get errors trying to run it on my Mac, need to try and figure it out

[@jeroenfrijhoff8392]

Update: had to turn off ‘Enable acceleration’ for the ‘AI hardware acceleration’ option in the ‘Advanced’ tab, then it works.

[@CuivTheLazyGeek]

Great!

[@SteveCollyer-m7j]

I get consistently better results for sharpening and denoising with Cosmic Clarity

[@user-yh8ce7tg8l]

Just looked into Cosmic Clarity, their satellite trail remove released yesterday (1-08) just sold me.

[@CuivTheLazyGeek]

Good to know, I'll need to compare them!

[@simongeorge2505]

Sorry, new to this so a really simple question, do you run deconvolution pre or post stretch? Also would it be better to use starless images? Thanks in advance.

[@CuivTheLazyGeek]

Pre-stretch! And no, deconvolution works based on a point spread function (PSF) that is deduced from the star shapes - so it would not perform correctly on a starless image!

[@simongeorge2505]

@@CuivTheLazyGeekThanks very much 👍

[@xyzzy7145]

consider test StarXterminator together with GraXpert Star decon

[@TheSpectralArtisan]

I am ecstatic to see there is a Mac copy… work couldn’t finish sooner lmao it’s 2:15… 130am let’s go let’s go lmao

[@CuivTheLazyGeek]

Hope it works well for you!

[@eMPHA5ER]

Me, watching your video 10:35 and starting laughing out loud by accident. How crazy is this result?

[@CuivTheLazyGeek]

Hahaha that's awesome :)

[@whyf16uy]

You didn't do any denoiseing in graxpert?

[@CuivTheLazyGeek]

No, I was comparing the deconvolution on its own - I have other videos where I compare the denoising.

[@TheCongratulationsChannel]

Is the dark ring-shaped area around the Plejiades real or an artefact of some kind? And I won't buy any astro software atm, as I have bad wether for weeks now and my Dwarf 3 has not arrived yet (after 13.5 weeks).

[@CuivTheLazyGeek]

Artifact!

[@GrouchoDuke]

Great vid as always, Cuiv. Thanks for spreading the word - don't use Honey!!

[@CuivTheLazyGeek]

Cheers mate!

[@TheDartFrog]

doesn't quite reach blurx levels, but for being free it's really good!

[@CuivTheLazyGeek]

Absolutely!

[@frankwakefield8815]

Can you run GRAX twice on the image and get even closer to BlurX?

[@CuivTheLazyGeek]

I personally wouldn't based on the difference of noise handling

[@davidletz9123]

"Pretty good and free" works for me.

[@CuivTheLazyGeek]

Yup!

[@dominickzaucha]

Blurx will be hard to dethrone. Its so good sometimes i wonder if its making up data.

[@CuivTheLazyGeek]

I tend to trust Russ's description of its implementation - and the fact that effectively the other decon algorithm discover a lot of the same details is a good sign!

[@davyhay1]

Do you need to uninstall the old version first?

[@CuivTheLazyGeek]

I installed over the old version as far as I can tell

[@DonnyHooterHoot]

I am always afraid that AI will "cheat" by identifying which area of space you are recording, getting clearer pictures from databases, then displaying the results to us. How would we know? Peace!

[@ciliiyou]

I asked AI to create me different astro image every night... More convenient than astrophotography... 😅

[@benjamincasatimcintosh2918]

"which made the hobby pay to win" my brother in christ you are doing astrophotography

[@BenCo-l3u]

Yeah baby

[@ulismoon]

money wins again...

[@rtyyyyb]

when wasnt astrophotography not pay to win

[@deepskytrekkin9600]

It's like wine . . . there's a reason why the good stuff is expensive. Just sayin'

[@MichaelZ-1]

Nice video Cuiv! Thank you for this insight! Just wondering though, here you process high quality taken images. What would comparison be if you took a smaller, cheaper camera and process that with both, and then compare between blurxterminator, graxpert and a high quality version of that place? For example, those green stars, maybe they are green and blurxterminator made things the wrong color and other details sharper or blurrier compared to what it really is. Comparing the images with a higher resolution photo with a better telescope, might give insight in that. (Or maybe I’m overthinking this)

[@CuivTheLazyGeek]

Thanks for the compliment about high quality images (taken from Tokyo!) - also green stars don't exist (look it up!)

[@kennethholmedal9694]

You are pixelpeeping now Cuiv. 😅. How will the whole image appear? 🤓

[@syberand]

$100 is not expensive in the world of astro

[@CuivTheLazyGeek]

True, but it does require PixInsight :D

[@charliemiller3884]

I just spent $15K on an imaging system. A hundred bucks for BlurXTerminator is a no brainer :)

[@CuivTheLazyGeek]

Hahaha relativism! :)

[@alanhnatiw1643]

I disagree with your stance that this is a pay to cheat. Is it cheating using a $10k mount vs. a $1k mount or using a $2500 cooled camera vs an iPhone…we are all amateurs just enjoying our hobby….given what most of us spend on hardware, $100 on great software is a bargain. I typically get an SNR improvement of 6dB or more….that would require 8x more integration time or 4x the aperture diameter. I think that is amazing value IMO.

[@chrislee8886]

Cuiv said “pay to win” and that’s how i see BXT as well.

[@alanhnatiw1643]

I agree, but it came across as a very “negative” thing to do….i don’t think there was any malice in what he said, just how I felt

[@CuivTheLazyGeek]

I only said "pay to win", that doesn't imply cheating, and it was said facetiously! Sorry about that, it wasn't intended as a serious thing or any indictment of processing techniques (that I myself use)

[@GusRobinson-MiserableOldGit]

I am fairly sure my mum wont care what Graxpert says about her.

[@justklaas4703]

Mmmm. I will wait for the non-beta version. Can't have AI cursing my mother.

[@CuivTheLazyGeek]

Hehehe

[@paulsmaglik5063]

The blur exterminator rocks, I am not impressed with GraXpert

[@MrGuilletv]

I Am , considering Its free... A year More and win

[@raphaelboisard7228]

It's a free Beta version so yeah it's actually quite impressive. And blurx is 100€+pix price so that's almost 400 euros

[@richardwills6930]

I'm wondering if Rx astro is going to regret going the pay to win route with PI and not bringing out a standalone, as free is fast catching up

[@CuivTheLazyGeek]

GraXpert is technically infinitely cheaper than BXT though! I think it's doing quite well for free - but indeed BXT is better!

[@esseremi]

how to destroy astrophotography

[@CuivTheLazyGeek]

How so?