AstroBin is the home of astrophotography - Share full resolution images, preserve acquisition details, discover equipment, discuss techniques, and learn from the community.Discover AstroBin

Weird gradient I can't remove through PixInsight (M31)

11 replies356 views
Elena avatar

Hello, this is my second attempt at using PixInsight, so I have processed the M31.

I found a gradient on each side of M31 and ran it through GraXpert and even increased smoothing but I found it couldn’t be removed?

image (1).png

This image attached shows the gradient on each side through range selection

Untitled130_20260718205528 (1).jpeg

This image shows my final & processed M31 attempt with the gradients still moderately evident?

If anyone might know what the issue could be, I would deeply appreciate some advice or help

Helpful Respectful Concise
SemiPro avatar

Sometimes GraXpert is not the best choice. When processing objects that taper into the background, it’s easy to create a ringing effect because of incorrect background extraction. Or, something else is going on. What does the original stack look like with STF before processing?

Well written Engaging Supportive
andrea tasselli avatar

Try MultiscaleGradientCorrection. Otherwise DBE with carefully chosen background sampling.

Elena avatar

SemiPro · Jul 19, 2026, 10:34 AM

Sometimes GraXpert is not the best choice. When processing objects that taper into the background, it’s easy to create a ringing effect because of incorrect background extraction. Or, something else is going on. What does the original stack look like with STF before processing?

I think so- this is the image after only ABE and SPCC for reference

image (2).png

SemiPro avatar

andrea tasselli · Jul 19, 2026, 10:42 AM

Try MultiscaleGradientCorrection. Otherwise DBE with carefully chosen background sampling.

I second this.

Elena · Jul 19, 2026, 10:52 AM

SemiPro · Jul 19, 2026, 10:34 AM

Sometimes GraXpert is not the best choice. When processing objects that taper into the background, it’s easy to create a ringing effect because of incorrect background extraction. Or, something else is going on. What does the original stack look like with STF before processing?

I think so- this is the image after only ABE and SPCC for reference

image (2).png

I wouldn’t even run ABE on it, personally. I think MultiscaleGradientCorrection or DBE like was suggested above will get you where you want to go!

Elena avatar

SemiPro · Jul 19, 2026, 11:06 AM

andrea tasselli · Jul 19, 2026, 10:42 AM

Try MultiscaleGradientCorrection. Otherwise DBE with carefully chosen background sampling.

I second this.

Elena · Jul 19, 2026, 10:52 AM

SemiPro · Jul 19, 2026, 10:34 AM

Sometimes GraXpert is not the best choice. When processing objects that taper into the background, it’s easy to create a ringing effect because of incorrect background extraction. Or, something else is going on. What does the original stack look like with STF before processing?

I think so- this is the image after only ABE and SPCC for reference

image (2).png

I wouldn’t even run ABE on it, personally. I think MultiscaleGradientCorrection or DBE like was suggested above will get you where you want to go!

Thank you so much for all the help!

Tony Gondola avatar

I would vote for multiscale with everyone else. Also, if you have access you can also try the new script in Siril called auto gradient removal. This has been my goto lately for dealing with complex gradients and sometimes it can get the job done when nothing else works. I guess the bottom line with gradients is there isn’t one tool that works for every case so you really just have to try different approaches and see what works.

Helpful Concise Engaging
James avatar

M31 has a lot of faint dust and star light around the main disk. If the various tools you’re using for gradient extraction are using this area as background reference points it could lead to this dark ring around the galaxy.

Thake a moment and look at some of the recent (past few years) M31 IOTD shots. Ignore the Ha stuff for now, see if you can find a few that show the extent of this region.

Then use Dynamic Background extraction and be a bit selective on where you are putting the reference points.. using the IOTD reference shots as a guide to help you determine where the actual back ground is.

Well written Helpful Respectful Concise Engaging Supportive
Ross Hrubiak avatar

Have you tried using the stand-alone GraXpert app? There you can manually select points that are background, which gives you more control than the GraXpert module inside PixInsight. You'd need to save the image as 32-bit FITS (or TIF?), open it in the standalone GraXpert app, export it there, and re-import back into PixInsight. I found this can produce much better results.

Well written Helpful Respectful Concise Engaging Supportive
Elena avatar

Ross Hrubiak · Jul 20, 2026, 01:00 AM

Have you tried using the stand-alone GraXpert app? There you can manually select points that are background, which gives you more control than the GraXpert module inside PixInsight. You'd need to save the image as 32-bit FITS, open it in the standalone GraXpert app, export it there, and re-import back into PixInsight. I found this can produce much better results.

I have not, I will try this- thank you so much!

Elena avatar

James · Jul 20, 2026, 12:07 AM

M31 has a lot of faint dust and star light around the main disk. If the various tools you’re using for gradient extraction are using this area as background reference points it could lead to this dark ring around the galaxy.

Thake a moment and look at some of the recent (past few years) M31 IOTD shots. Ignore the Ha stuff for now, see if you can find a few that show the extent of this region.

Then use Dynamic Background extraction and be a bit selective on where you are putting the reference points.. using the IOTD reference shots as a guide to help you determine where the actual back ground is.

Thank you so much for the help, I will definitely try this!

Well written Respectful
Jonathan Paul avatar

DynamicBackgroundExtraction should fix that.