Back to forums

Gradient TOOLS compared: GXP, GC, ABE

Pleiades Astrophoto PixInsight Steffen Hirtle GraXpert Processing Gradient Removal Deep Sky
8 replies•681 views
Aleix Roig avatar
Aleix Roig avatar
Good morning,

Not pretending to share a deep analysis, just a quick test on some good data that I captured from Bortle Namibian sky last July 2024 (Tivoli Astrofarm). Image details: L = 96x300" (8h) ; FRA300 + ASI2600MM + AM5).

I wanted to test the results of several Gradient Correction Tools. What I did with the data (all in PIXINSIGHT):

-Dynamic Crop
-BX
-Test of different Gradient Correction Tools as it follows: ABE (PIX), Gradient Correction (PIX) two times (v1 is default settings; v2 I lowered Scale/Smoothness and ticked Automatic Convergence... it finished after 4 iterations), GraXpert (standalone).
-SX to see the results

What are your thoughts on the results?

My humble opinion:

-ABE (PIX) overcorrects several areas, not doing a great job on several corners. I used it just to compare, knowing its limitations.
-GXP overcorrects the data way too far; introducing some extra noise (?) 
-Gradient Correction (PIX) does a great work with a little fine tuning of the settings

Just to take into account... The images are still linear and I auto stretched it... so the noisy view may lead to confusion. I just wanted to compare the Gradient Tools, not the noise introduced (but it'll also be interesting to see its effects).

Conclusion: in this example I prefer a non AI based gradient reduction tool, GC (PIX). 

Thanks for reading and for your thoughts. Best wishes and clear skies to you all,

Aleix Roig  
​​​​​​​www.astrocat.info

Helpful Insightful Respectful Engaging
Aleix Roig avatar
Aleix Roig avatar
Just the final results stretched again:

GraXpert vs Gradient Correction

Fritz avatar
Fritz avatar
Aleix,

thank you for sharing this one. There is one more: DBE. In many cases, this is my favorite, as it gives me best control. But it needs more effort to use, as all depends on placing of the patterns. This involves multiple runs and often even some investigation of the FOV to distinct structures from background.

CS
Fritz
Aleix Roig avatar
Aleix Roig avatar
Fritz:
Aleix,

thank you for sharing this one. There is one more: DBE. In many cases, this is my favorite, as it gives me best control. But it needs more effort to use, as all depends on placing of the patterns. This involves multiple runs and often even some investigation of the FOV to distinct structures from background.

CS
Fritz


So true Firtz, I skipped DBE and it’s a great tool. Much better than ABE (IMO).

Best regards,


Aleix
Monty Chandler avatar
Monty Chandler avatar
GC has proven to be an awesome tool.  If it can't do the job, then I go back to DBE and spend the time there.  It's rare that I have to use DBE these days.
Bill McLaughlin avatar
Bill McLaughlin avatar
I have been using DBE for years and more recently Graxpert (in non-AI mode, I almost never use the AI option).  I have also tried GC and found that it simply fails to do as good a job as Graxpert or a careful use of DBE in the majority of cases. GC settings are finicky and in less time than I would spend doing that  I can get a better result with either Graxpert or DBE.

IMHO, there is yet no good substitute for manually placing sample points and  automated placement should just be used as a starting point. One reason I like Graxpert is the ease with which manual placement and rearrangement of points can be done.
Helpful Insightful Respectful
Rodolphe Goldsztejn avatar
Rodolphe Goldsztejn avatar
Thanks for this benchmark.

I usually start with GraXpert to get a first impression about how the gradient looks like, on a copy of the image.
Then play with GC on a preview to obtain a similar result, which often means fine tuning several parameters.

I’m not always successful with this approach to be honest, especially on fields filled with nebulosity; I may thus just skip this step.

Eventually I decide which result seems the best to my eyes, also considering how noise reduction would finish cleaning up the background at a later stage of the workflow.
Helpful
jmarinotero avatar
jmarinotero avatar
Aleix,

Good comparison, this matches the tests that I have conducted on my own data. GraXpert can work wonders, but on some images it just fails. I believe GC is a great tool, but it is a bit deceiving in its simplicity and it's certainly not a "one button push" kind of tool
Christian Bennich avatar
Christian Bennich avatar
Nice simple comparison. 

I still test DBE, GraXpert and GC against each other when I do gradient removal. 
I typically end up with the version from GC after fiddling with settings.