Pixinsight GRADIENT CORRECTION - Does it keep its promises?
Vložit
- čas přidán 29. 02. 2024
- Today PixInsight released the next gradient correction process, which makes ABE and DBE redundant. But is it better than GraXpert AI and hence is it the new astrophotography champion in Gradient removal? I did pressure test it with some extreme pics and compared it with GraXpert. You will be surprised!
Join my Patreon site for getting cutting edge news about Astrophotography software and equipment, early access without commercials and tons of supporting documents: www.patreon.com/user?u=83665984
If you buy any equipment you may consider these three shops - by using the links below you support the channel:
Agena Astro: agenaastro.com/?rfsn=7982331....
High Point Scientific: www.highpointscientific.com/?...
Astroshop.eu: www.astroshop.eu/?affiliate_i...
#astrophotography #pixinsight
------------------------------
Music credits:
ORBITAL_StriKe by B E T T O G H | / bettogh
bettogh.bandcamp.com | open.spotify.com/artist/3zDlq...
Music promoted by www.free-stock-music.com
Creative Commons Attribution 3.0 Unported License
creativecommons.org/licenses/... - Věda a technologie
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
Thanks for always keeping us up to date Sasha! Great comparison and overview ❤
Very impressive tool and I'm impressed how you evaluated the tool so quickly to help us. I mean it will be a few days before I have time to even do the PI update. Well done
An absolutely incredible addition to Pixinsight. Using it on a few of my older images and ones with gradients it's super impressive.
A new tool with great promise. I had a little play with it earlier, I was very impressed with it. If Pixinsight could add a live preview to the process, it will be even better.
Yes agree, that would be the cherry on top of the icing!
@@viewintospacein french we say « la cerise sur le gâteau » too 😊
Excellent tutorial
When I received the email from Pixinsight yesterday about the new version that features Gradient Correction, I said to myself, “I wonder what Sascha will think of this.” And this morning, here you are! Well done. Now to see what Luke thinks.
Hello Sascha, GraXpert does ONLY work correctly if you crop the picture first. If you did not crop the pictures first in your comparison it might be not fair. Just a comment from my side, regards from Austria.
I'm fully aware of that and I did crop all of the images I used. In one I forgot two corners and that had an effect but otherwise the comparison was fair. In any way, even this requirement to crop the pic is a limitation of GraXpert that the new process does not have any more - another big advantage.
You are absolutely right. But from the other side, that’s a +1 for PI if it can handle the gradient correction without crop.
This should also allow to crop less.
Brilliant!
Thank you for making this comparison so I didn’t have to field these type of comments as well 😂. Great review as always!
Many thanks for clarification an best regards from Graz.
@@viewintospace
I installed the new update last night and also went straight to some of my worst old images to try the new gradient correction process. And my reaction was pretty much the same as well, just wow, mind blowing. Not only does it work amazingly well, the speed was honesty the most shocking part to me.
I'm a big GraXpert fan and I used it almost exclusively over DBE, even before it added the AI interpolation method. And I think it still has some specific uses where it can replace or complement gradient correction, but for like 90% of use cases this new tool is clearly the way to go.
Thanks for the quick video .. Looks like some of my old images too can be resurrected now!!
Im seriously so excited to have this tool. Gradients have always been the thing that I have struggled the most with.
I appreciate your content! Thanks!
Just wow! Now I really need to try this on my bad Bortle 8 gradients.
Absolutely amazing
No more flats that’s game changing! Really well explained as well Sacha I will get downloading the new version
I would still do flats, but for the case you forgot it or have an issue with them, you have a backup!
I still feel you should do flats for dust but that's just me I mean doesn't take too long to do flats anyways but we shall see
@@caviestcaveman8691 Fully agree - they are still needed!
Good job, looks really impressive.
Can't wait to try it! But I think that taking flats is still going to be important. Flats will eliminate gradients inherent in your optical system, and those corrections will be specifically tailored to your system, doing so near perfectly. Gradient correction will do what flats cannot, taking care of external sources of gradients-light pollution, the moon, etc. But it's a more general solution that needs to be tweaked to get the best results for a specific image. You could use gradient correction for both, but using both flats and gradient correction should produce the best results.
100% agree!
My workflow has always been to crop artifacts and then make 3-4 clones of the image, then run ABE, DBE and GXT (and now GC) on each; evaluate each outcome, then move on. I’ve also found that the application of the gradient tools seem to work nicely on each of the separate channels, than recombine the corrected images. Not fast (or easy) but having choices and seeing differences early in the workflow always seem to pay dividends for me….
Great video. I would of liked to see the comparison first by using the new tool with defaults and then if not so good to play with the sliders.
I used it this morning on 2 pics in standard mode, as I wanted to try something out, and the results were worse that what I showed yesterday in the video. I have a strong feeling that this is not like GraXpert just pushing a button, but that the result really depends on understanding each of settings and applying it correctly. The approximate way I applied yesterday in the video gave already quite good results, but I will also watch the (at least 2 hour) video Adam Block will for sure at one point release, to fully grasp the exact mechanics of each of the settings.
I just did a quick comparison of gradient correction vs graxpert using just default values (except checked convergence on GC). I used one raw 3min ,uncalibrated Lum sub of NGC 2403, no cropping. The graxpert result was much better than the GC. The GX version was gradient and artifact free, but the GC had several unacceptable artifacts. Probably tweaking the parameters would help but I wanted to see the result right out of the box. I usually dont process uncalibrated, unstacked subs so this doesn't represent a typical work flow for me, but after all the hype, I was expecting more.
GC is not a one button tool - you have to adjust the settings to get good results, and the more experience we have and hence the better we adjust the settings, the better will be the result
undoubtedly true Sasha, nonetheless, I've not had any issues with GX and may turn to GC if I do, but if I can get great results with one button (GX), I won't bother with a tool that I need to tweak. I'm not a GX fanboy, I'll use whatever produces the best results.
Many thanks for this…. Very informative. Any idea when MARS might be released?
No clue, but given they are still collecting data this might still take a while
Thanks for the prompt review with comparisons. Very convincing data you presented. Are you convinced that flats are less relevant now?
No, flats are still needed. But at least if something goes wrong or they are missing there is a way to mitigate it.
Thanks Sascha. Graxpert is toast but what do you think? Gradient Correction or SetiAstro's Automatic DBE script or both? Would love to see a comparison rather than rely solely on my own judgment.
Sometimes it needs a comment like this to trigger me into action! Video is shot but still needs to be edited - most likely be released tomorrow. But if you want already my finding: Auto DBE is much worse then GC - no chance at all.... and I was not really surprised....
How do you get it . I use hyper star some times . Had bad flats on 30 hour's off Iris Nebula
Just upgrade (downloading the new version on the PixInsight homepage)
Hey Sascha, I have followed you in using GHS later in the work flow. Since the 1601 update, which btw crashed PI once, and after using GC I find my old GHS techniques result in a very green screen (I shoot OSC). The histogram shows shows substantially separated R, G, B peaks. Did something get reset so I always get unlinked results. I thought before they were more like linked results. Anyone having similar problems?
Have a look at my Toolbox video and specifically at the Auto-LinearFit script - this might solve your issue!
@@viewintospace I discovered Auto-LinearFit since I posted this comment. I think my mistake was not realizing that SPCC does linear fit, so GHS gets a different starting point if SPCC is done than if it is not. You and your videos are a great source of information.
But what happens when you re-stretch after using the gradient correction tool?
Not much... just tries it out and results stay 99% the same good
So, am I correct that this replaces DynamicBackgroundExtraction? There doesn't seem to be any reason to use DBE any more, yes?
As always, it is good to have option as the best tool might sometime fail. So GC is surely the standard process from now on, but for the very rare case that is does not achieve satisfying results it's good to know DBE and GraXpert.
I prefer DBE or even ABE over GraXpert. Haven't downloaded this version yet.
it seems to be a very nice addition, i know it’s to be used in conjunction with the MARS tool coming up but honestly i think it’ll be made irrelevant by the latter. but man i can’t wait for the MARS project to conclude, it will completely change the game for amateur astrophotography, seriously i think everybody who uses it will have leagues better images. as for whether flats are irrelevant or will be irrelevant, i would say no, as this tool at least seems to only be for additive gradients, and when you use subtraction to correct for multiplicitave gradients you often end up with wrong colours, especially if there are structures in the corrected edges. if this tool and the MARS tool end up allowing you to choose for what type of correction you need, and it does multiplicative and additive gradient correction separately, then yeah flats will be made irrelevant too :)
Did you mean to have the smoothing set to 0 on GraXpert for M14?
The developers kind of state that smoothing on GrXpert does not match the smoothing in DBE/GC and should as default always left at 0.
I wonder if GraXpert would do a better job on M100 if the image was prepared according to their specs. Afair you're supposed to crop the image so that no stacking artifacts remain. Otherwise GraXpert won't do a good job at all.
As for the video itself: Thamks for all the examples and some usage examples. They sure will come in handy.
For the M100 yes, I missed the 2 corners when I cropped and that surely affected the result. BUT: Overall in any of the cases the Gradient Correction did a better job, also with the pics which were cropped appropriately and we also know now that Gradient Correction is not affected by not well cropped pics, which is a further advantage of Gradient Correction. I was really a fan of GraXpert and I kind of wanted it to win, but the results speak a very clear language.
@@viewintospaceYes, I also pulled out some of my old images where I had problems with ring-like structures, supposedly caused by a lossy compression of my Nikon RAWs, and the new gradient removal tool did a wonderful job on them.
So it's definitely a new tool in my toolbox.
Why didn't you try to increase the smoothing in GraXpert?
Because the GraXpert team kind of tells you to leave this setting at default value. But besides that, there were worlds in the performance of these tools, even if the change of smoothing would have improved it a bit, it would have never achieved the results of the Gradient Corrector.
@@viewintospace I just updated to the latest Pixinsight version and tried out the new gradient tool on a few raw images. your right, it works fabulous!
I"m not seeing the new update yet?
Go to the Pixinsight Homepage and download the most recent version, then reinstall and then search for updates...
Does this obsolete backgroundNeutralization too? Seems like it does.
Nope, this is just gradient removal like DBE. But if you use SPCC Background Neutralisation is included there.
Not working . It gives me a message: *** Error: Insufficient data to sample gradient, channel 0 and failed. It also deleted it for me the starnet++.
You downloaded the new version from PixInsight.com, the updated in PixInsight everything and restarted it? If not sure, have a look at this video: PixInsight Troubleshooting: Common Issues and How to Solve Them
czcams.com/video/PfAIMIA-m3E/video.html
Try unclicking structure protection - worked for me on a Siril stacked difficult image.
@@andysmith7903 Thanks is working but i think this is bug .
Hello Sasha, good video BUT imiss a compare with nebula pictures ;) cs Mani
I agree, kind of got forgotten. You might watch the video of Lukomatico, I'm sure he has a nebula pic. But I would bet the Gradient correction will also work better there, as I had a few times issues with GraXpert that it also did eat away a chunk of the nebula.
Would be nice as Graxpert has murdered a couple of images of mine, used the APP pollution tool which worked well.
And when I said murdered - the LMC has a cloud of nebulosity around it and Graxpert just stripped it out. Same with Rho Ophiucus - the subtracted gradients were the same colour and shape as the nebulosity.
Would be interesting to see if the PI tool overcorrects in the same way.
You still need flats for dust on the sensor and lens. Hopefully very soon we won’t need anything but a computer and Pix AI. No flat, no dark, no bias, no light, no filters, no camera, no dithering, no scope, no dew heater, no mount, no nothing! Just imagination. Just kidding ! I’m a big fan of these new tools for astro !
You still need flats, that is correct - I did this also address in the follow up video. But regarding dust - it was quite amazing how GC did remove the doughnut clearly origin from dust in one picture I show here.