PixInsight New Tool - Multiscale Gradient Correction Test [Deep Sky] Processing techniques · Krzysztof Gaudy · 12/21/2024 · 4 · 607 · 8

Yesterday a new version of PixInsight 1.9.0 was released. The release contains some nice features, but by far the most important one is the MultiscaleGradientCorrection tool (I):  https://www.youtube.com/watch?v=BHE3J069NCo

It is supposed to do background gradient correction, of course Today I spent a couple of hours with this tool and tested it. I got some of my old Bortle 4 data which I know has some serious gradient problems. To my surprise, I found that the default settings did not work very well. I had to spend some time testing it, but I think the end result is impressive. 

Before (print screen of stretched unlink STF just after Channel Combination). Green gradient on the left, green gradient in the middle, magenta on the right and blue on the bottom... 


And after MultiscaleGradientCorrection:


Very nicely removed, especially the green gradient. I used DBE then + SCNR 100% green for some final background neutralization and got the result, which is great in my opinion. All gradients probably removed, at least at this stage of image processing. 


The settings that work best for me after many tests are: 

Gradient Scale: 192 - which is significantly different from the default but it certainly depends on your equipment. 
Structure Separation: 2
Model smoothness: 2
The rest are defaults. 

After some final adjustments in Photoshop and LRGB combination I think I was able to achieve some decent result: https://www.astrobin.com/eym0ts/

My opinion? It's not a revolution, probably with some advanced DBE we should get the same effect, but it certainly shortens the whole photo processing. 

If you want to try processing the above image yourself: https://astrophotomarket.com/en/image-details/30/barnard-150-the-running-shadow - here are master lights files available for free download. Maybe you will find better settings and results!
SpectrometricFlux settings for the data:


I'm curious what your feelings are after testing this new tool and whether it will be included in your workflow?

Krzysztof Gaudy:
Yesterday a new version of PixInsight 1.9.0 was released. The release contains some nice features, but by far the most important one is the MultiscaleGradientCorrection tool (I):  https://www.youtube.com/watch?v=BHE3J069NCo

It is supposed to do background gradient correction, of course Today I spent a couple of hours with this tool and tested it. I got some of my old Bortle 4 data which I know has some serious gradient problems. To my surprise, I found that the default settings did not work very well. I had to spend some time testing it, but I think the end result is impressive. 

Before (print screen of stretched unlink STF just after Channel Combination). Green gradient on the left, green gradient in the middle, magenta on the right and blue on the bottom... 


And after MultiscaleGradientCorrection:


Very nicely removed, especially the green gradient. I used DBE then + SCNR 100% green for some final background neutralization and got the result, which is great in my opinion. All gradients probably removed, at least at this stage of image processing. 


The settings that work best for me after many tests are: 

Gradient Scale: 192 - which is significantly different from the default but it certainly depends on your equipment. 
Structure Separation: 2
Model smoothness: 2
The rest are defaults. 

After some final adjustments in Photoshop and LRGB combination I think I was able to achieve some decent result: https://www.astrobin.com/eym0ts/

My opinion? It's not a revolution, probably with some advanced DBE we should get the same effect, but it certainly shortens the whole photo processing. 

If you want to try processing the above image yourself: https://astrophotomarket.com/en/image-details/30/barnard-150-the-running-shadow - here are master lights files available for free download. Maybe you will find better settings and results!
SpectrometricFlux settings for the data:


I'm curious what your feelings are after testing this new tool and whether it will be included in your workflow?

Looks good, but in my testing, the older tools do just as good a job as the new MSGC, DBE or GraXpert, or even better Auto DBE by SETI Astro…

Is MGC revolutionary?  Definitely not.  There already exist tools based on the same ideal (using a wide-field, gradient free image as a reference).  The difference is only implementation and ease of use.  If it's not the first one, it can't be called revolutionary (unless I misunderstand what revolutionary means in English, which is possible, I admit).  Based on everything I have seen thus far, I would vote for MGC for better implementation and easier to use.

But if you ask whether the general idea is a game changer or revolutionary, I suggest to conduct a much more challenging test.  Find a picture where bright and faint nebulas are everywhere, essentially filling the entire image area. On such an image, run MGC (with MARS database), and all other gradient tools you can find.  Then compare the results, better to compare with a ground truth, to see which one gives you a result close to the ground truth.