BlurXterminator technique and usage thread

Chris White:
Starting a new thread for folks who want talk about using this tool.  How have you used it, what tricks have you learned.  Have you found any applications where it was useful other than what it was intended for?  Where in your workflow are you inserting it?  What don't you like about it, pitfalls?

I've got some ideas to contribute and as I do some more testing I will add my thoughts and what I have learned. 

If you want to discuss ethics or anything else off topic from using this new tool... please find somewhere else to do so.

On Dec 17, 2022, at 10:47 AM, RC-Astro BlurXTerminator Support <[email]support@rc-astro.com[/email]> wrote:
\"When combining narrow-band data into a color image, it is best to avoid \"mixing\" color channels, as is done in some advanced color palette techniques, prior to deconvolution. If channels are mixed prior to deconvolution, especially with very different gains, star profiles can be significantly altered, perhaps so much as to make them unrecognizable as stars to the neural network. Do a simple SHO color combination, run BlurXTerminator, then perform any mixing between channels afterwards.\"\r\n\r\nFrom this, I am assuming that this implies combining the linear S, H and O masters, running BXT, then separating the channels for further steps, including blending these \"primary\" sources for each R, G or B channels (using the BXT-processed S, H and O in, say, R= .75Sii + .25Ha, where Sii and Ha are now BXT-processed as above). I would be grateful if you could confirm or correct my interpretation of the excerpted documentation from BlurXterminator.

Chris White:
Starting a new thread for folks who want talk about using this tool.  How have you used it, what tricks have you learned.  Have you found any applications where it was useful other than what it was intended for?  Where in your workflow are you inserting it?  What don't you like about it, pitfalls?

I've got some ideas to contribute and as I do some more testing I will add my thoughts and what I have learned. 

If you want to discuss ethics or anything else off topic from using this new tool... please find somewhere else to do so.

Brent Newton:
Workflow wise is simply replaces Deconvolution in most or all cases so far. I've been tearing through several of my previous shots lately and so far have not found a single deficiency in use of the process. At worst, it may require a manually-set PSF value (and I suggested to RC that it may benefit from being able to accept sample PSF images as the Decon process does now) but this is barely a complaint. Many of my DSLR lens-based integrations or my older Doublet work suffered/suffers from poor corner stars (imperfect focus or loose focus tubes) and while I previously had to crop out what I considered to be unfixable areas of an image, BxT performs only what I can describe as a miracle.

Alex Ranous:
Has anyone used the tool on a Mosaic yet?  I'm wondering if it's best to use BXT on the individual panels before combining, or after it's been assembled.  As I'm lazy, I'd prefer to do it after, but I'm wondering if there are benefits after.  If your edge stars are problematic, then BXT may correct them, improving the result.  I guess I'll have to do it both ways and see.

Alex Ranous:

Aygen:

Alex Ranous:
I just posted my first images using BXT.  I picked m8 as I had issues when I first processed the data this summer, so I shelved it.  I shot it with my focal reducer, so at a pixel scale of .89 arcsec/pixel, which was more amenable to deconvolution than 1.3 when I'm using my reducer.  I found the defaults a bit too strong.  I ended up dialing back to about .8 for the non-stellar features and about .2 for the stars.

As I didn't like my original version due to other processing sins, I don't have a comparison.

RGB version of M8

SHO version of M8

Has anyone used the tool on a Mosaic yet?  I'm wondering if it's best to use BXT on the individual panels before combining, or after it's been assembled.  As I'm lazy, I'd prefer to do it after, but I'm wondering if there are benefits after.  If your edge stars are problematic, then BXT may correct them, improving the result.  I guess I'll have to do it both ways and see.

Alex, your images are great ! Can you give some details about your workflow ?

So the raw subs were processed with WBPP with all the normalization stuff.  For this target I didn't do any gradient stuff.  There was so much nebulosity everywhere I couldn't tell what may have been gradient vs nebula.  For the RGB image, I combined RGB and color corrected using SPCC.  I then hit it with BlurXTerminator.  I ended up backing off the non-stellar sharpening to about .7 or .8 if I recall.  I believe I kept the default star shrink.  

I then extracted the stars with StarXTerminator and stretched both the starless image and the stared image separate, the stars up to the size I wanted.  I then hit the RGB with NoiseXTerminator and various curves adjustments for saturation and brightness tweaks.  For the Lum I hit it with BXT with the same settings and the RGB, and then removed the stars.  I stretched it and in addition to curves tweaks, did some LocalHistogramEqualization and the DarkStructureEnhance. After the LRGBCombination, I combined back the RGB stars with the op_screen operation in PixelMath.

I do SHO a bit different.  I used BXT on each individual filter, then tossed the stars.  I stretch each individual filter separately and tweak them individually in curves to get the levels ok for the SHO channel combine.  I then blended my Ha and SII (about 70/30) to create a lum.  I then did some LHE and DarkStructureEnhance to get it looking the way I want.  I used some masks to selectively tweak different regions also.  I believe I also used the EZ Processing Suite's EZ HDR script to tone down some of the really bright areas.  I then LRGBCombined using that new lum with my SHO image, and added in the stars from my RGB image.

I tend to do a lot of tweaking using curves and sometimes masks. so a lot of what I do is very situational.  I often use the GAME and ColorMask scripts to create masks so I can target specific areas for work.

Hello,
I asked Russ about when to use the tool on SHO data, whether to run first on the individual stacks of mono data, or combine and then run BXT, and he told me to do the later, because I would probably varying FWHM figures across the different stacks, so best to do it this way….🤔🤔