How to use BlurXTerminator with AutoIntegrate?

I’m not sure if you have seen the newly introduced BlurXTerminator (BLX) from rcAstro yet?
It seems truly remarkable.

In this video at 7:15, Adam Block recommends that BLX be used while the data is in it’s linear state right after the creation of the RGB image…it’s ok to use DBE and Color Calibration but that seems about it.

I was wondering if it is possible to use BLX in some sort of 2 step process in version 1.54?
i.e firstly use autointegrate to process up to the creation of the RGB image with only DBE and Color Calibration applied.
Then manually run apply BXT to the RGB image.
Then restart autointegrate from a certain point to perform any additional processing.

My apologies, I am only a relative beginner (3-6 months).
I purchased PixInsight and was completely overwhelmed and ready to give up.
Then I discovered autointegrate.
Your product is solely responsible for the joy that I get from astroprocessing.
I am truly grateful.

Hi Garth,

I just downloaded BlurXTerminator two days ago. I agree, it looks very good.

I have a version of AutoIntegrate in GitHub (GitHub - jarmoruuth/AutoIntegrate: PixInsight script to process FITS and other image files) that optionally uses BlurXTerminator for sharpening. But it uses just defaults (no configuration yet) and runs it quite late in the process for non-linear image. I will add it to an earlier spot in the process while the image is still linear.

I have not tried it myself but you should be able to apply it manually to Integration_L + Integration_RGB images, rename those to Integration_L_start + Integration_RGB_start and use AutoContinue (AutoIntegrate Info). For OSC, you would have only Integration_RGB.


1 Like

GitHub now has a version that runs BlurXTerminator on the linear data.


1 Like

Hi Jarmo,
Thanks for getting back to me.
Very much appreciated.

That’s fantastic. Thanks so much

This don’t have much to do about the new BlurXterminator, but you mention about being new to PI and I know how over whelming it can get at times. I’ve been using it for about 2 years now. But I found a youtube chanel that may be of some help to you. View Into Space, He does a good job of explaining each modual/process you may want to check it out. It’s helped me to understand some of them.

1 Like

Hi Lostone,
Thank you.
I will definitely be checking these out.

I was using astropixel processor initially.
It’s very good at what it does and reasonably easy to use, but doesn’t have the breadth of PI.

PI is next level but has a very steep learning curve for someone at my level.

Finding autointegrate was like striking gold. I could produce vastly superior images using PI without the steep learning curve. I am so grateful to Jarmo.

I think I’m at the point where I’m curious about the workflow steps and tools used in each stage.
I’m sure these videos will help as I start this journey.
Thanks again

Hi Jarmo,

I was reading through posts in the Adam Block Video and noticed this reply from Russ Croman in response to how where to use BXT in a narrow-band workflow.
Here is RC’s response:

“For NB, I would 1) channel combine linear to SHO, 2) calibrate color using SPCC NB mode or ColorCalibration with nebula as white reference (both are described in the SPCC doc), 3) run BXT. After BXT decon, if other palette/stretching choices are desired, extract back to separate channels and have at it.”

Just posting it in case it’s useful.
It’s way over my head but if you are wanting to do this in the future then it may help?

Hi Garth,

Thanks for the note.

Those steps are basically the same that are done in AutoIntegrate by default. But my understanding is that color calibration does not make much sense with narrowband data so it is not done by AutoIntegrate when using NB data.

I am not sure if you need to use SHO for BXT, maybe just using real channel data like OHS or HOO would be good too. What I was reading was to avoid mixed channel combination like “0.3H + 0.7S”. Support for mixed channel combination and BXT would need some changes because that would need steps to 1) combine SHO 2) run BXT 3) separate channels 4) recombine with user combination.

Right now BXT is done just before stretching. By default noise reduction is done for each channel. My understanding is that noise reduction should be done after BXT. In AutoIntegrate there is an option Non-linear noise reduction which moves noise reduction after stretching. So that could be good with BXT. In my tests I was not sure if that is better for all data, though.

I may push a version with BlurXTerminator out soon for automatic update although I am not fully sure if BXT is used correctly. And then there is support for SPCC but it looks so complex that I have not even tried it.


Hi Jarmo,
I can see that you published 1.55 with BXT.
That’s great.

I have been testing various combinations on M51 and I have noticed that using Noise XTerminator in the AutoIntegrate process produces a poor image.

I am using: Crop to common area, Use ABE on combined images, StarXterminator, and BLX and the image is reasonably good although seems to have poor color definition (very Yellow) compared to an image processed without BLX (which has nice red colors which are completely absent in the BLX image).

I’m not sure if I am doing something wrong but I prefer the AutoIntegrate image without BLX.

I should mention that I am using AutoIntegrate v1.56 test 1

Hi Garth,

Thanks for the feedback. I may have seen something similar even without BXT but never looked it more closely. If it is more visible with BXT I need to run my refence test set with BXT and try to see what happens there. My testing is unfortunately somewhat limited so I am happy to get feedback on any problems.

Are you using mono or OSC camera?

v1.56 test1 is similar to v1.55 in processing.


Mono LRGB from TelescopeLive.
I am only using 4 from each channel so that I can process it fast.

I have just run AutoIntegrate without BLX, then run the PSFImage script to get the FWHM values to enter into BLX. I run BLX on the AutoLRGB with the manual PSF Diameter value and set Sharpen nonstellar to 0.5 (instead of 0.9) and it looks ok to me (although I probably don’t know enough to know if I have done more harm than good).

Just ran M51 without “Use ABE on Channel images” and selected: Crop to common area, SXT, NXT and BLX and the image has turned out better.

So it must have been something to do with ABE on Channel images and NXT together?

Thanks for the tip! I have not used ABE early in the process much myself as I have not seen any benefits (Telescope Live). But I see many people use ABE (or DBE) early in the process so I need to recheck that. What I have read is that ABE should be ok for BLX.

Some things are a still bit unclear to me. For example AutoIntegrate uses linear fit and I am not sure if that is bad for BLX. Maybe I need to create RGB for BLX only and after that separate channels for further processing like linear fit, channel ABE, channel noise reduction etc.

Just to make sure, you are not running SXT before BLX? I understood that BLX uses stars to automatically get PSF.


1 Like

No, didn’t use STX before BXT (the image included stars and galaxy). I watched a YouTube video yesterday that pointed me to PSFImage.js Script to calculate FWHM values. The author felt that this provides a more accurate value than BXT…I did notice a slight improvement. If I was to use SXT then I would have run the PSFImage script on the star image and entered that value into BXT for both the star and starless images.

Also, I tried Non-linear noise reduction.
It produces a strange blotch in the center of the galaxy. It’s just a blob of a single color that doesn’t blend into the rest of the galaxy…like there’s no feathering if that makes sense. I think it’s messing up the masking somehow?

Right now, I am getting good results from AutoIntegrate with the following settings:

  • Crop to common area
  • Use BackgroundNeutralization
  • SXT, NXT and BXT

Also feel that I’m getting better results selecting “No noise reduction” then applying NXT on the autoLRGB image.

When I do select NXT, I can see it running before BXT in the console window. I’m not sure if that’s good for BXT?

So my current best AutoIntegrate settings are:

  • Crop to common area
  • No noise reduction
  • Use BackgroundNeutralization
  • SXT and BXT

Then run NXT manually on autoLRGB

With option Non-linear noise reduction the noise reduction is done after BlurXTerminator and after stretching. It is recommended to run NXT after BXT.


1 Like

One of my test cases is Trifid Nebula. I can see that using BXT generates a slightly different image. But it is only because AutoSTF stetches it more so it is a bit more bright. This must be because of smaller stars in BXT image. If I make it darker with HistogramTransformation it looks very much the same. I used Non-linear noise reduction in my test and I did not use NXT.

Need to try with M51 next.


My experiments with M51.

Here is BXT with non-linear noise reduction vs default. Looks like BXT has less blue.

But when I compare BXT to default processing with non-linear noise reduction (no BXT) they are more close in color. Here the only difference is BXT vs normal sharpening.

So maybe the default of running noise reduction on each channel image makes the colors slightly different. Also details are better when using non-linear noise reduction. Channel noise reduction may not be the best choice for good data like we have here.


1 Like