BlurXTerminator - Wow!

[blinky]

Oh, thanks Olly - just installed both trials to see if they worked, now need to find $150 to buy them as they are so good!

[ollypenrice]

Humph, I'd like to be trying it, too, but I updated to the latest PI - or tried to - and they are not sending me my new activation code, or not in the 'few minutes' promised.

Olly

[geeklee]

It's going to be an interesting tool to experiment with.  I've found using deconvolution in PI previously at longer focal lengths has gotten me a subtle but noticeable benefit.  

I've tested on some recent data at medium FL - 675mm - and found again a noticeable benefit for sure, certainly not in any concerning way (i.e. anything made up).  My data was average it it managed to make it slightly less average .  I have seen some images going about where the "after" result is quite staggering in the difference.  Like a lot of processes (all?) good data in should mean good data out!

It's the star size reduction and sharpening that was much more obvious and probably welcome in some cases.

Here are a couple of my short FL examples at full resolution (they are crops).  RedCat + 294MM + 3.5nm Ha.  Left original, right BlurX.  

These are a touch too strong perhaps, but for IC443, the original looked good anyway... for 250mm (3.8"PP)!

Edited December 16, 2022 by geeklee

[Budgie1]

41 minutes ago, ollypenrice said:

Humph, I'd like to be trying it, too, but I updated to the latest PI - or tried to - and they are not sending me my new activation code, or not in the 'few minutes' promised.

Olly

I bought BlurX last night and haven't received the activation code email yet. I did get one signing me up to the updates email list but nothing else. 

[Stuart1971]

52 minutes ago, Budgie1 said:

I bought BlurX last night and haven't received the activation code email yet. I did get one signing me up to the updates email list but nothing else. 

Me too, that’s all I got so far….well this is just for the trial version….

Edited December 16, 2022 by Stuart1971

[tooth_dr]

I bought it earlier and got the license code 7 minutes after buying it.  I’m glad I didn’t process much data over the past year given that this has to be done at the start 😂

[gorann]

4 hours ago, ollypenrice said:

That's the wrong way round. You should run BlurX first (Adam Block explains why in the video.) Good result, even so.

Olly

Yes, but what confuses me is that the examples Block shows seems to be applying it to stretched and almost finished mages.

[tooth_dr]

4 minutes ago, gorann said:

Yes, but what confuses me is that the examples Block shows seems to be applying it to stretched and almost finished mages.

I didn’t think so, he explains he’s only had it a few days so he doesn’t have that many examples to show as he had to reprocess his data.  I have only watched the first 30 minutes but he doesn’t just apply it to almost finished images. 

Edited December 16, 2022 by tooth_dr

[gorann]

18 minutes ago, tooth_dr said:

I didn’t think so, he explains he’s only had it a few days so he doesn’t have that many examples to show as he had to reprocess his data.  I have only watched the first 30 minutes but he doesn’t just apply it to almost finished images. 

I saw the whole video, but maybe I got it wrong. However, near the end he demonstrates the artifact that it can produce, exemplified by the Crab nebula. What worries me then is that if you apply it early on and then much later realize that it has produced artifacts, it will be a bit frustrating. It suggests that you need to do two processes, one with and one without BXT.

[Spongey]

I just got the trial version to see what the fuss is about. I ran it on an image of mine that I captured around this time last year. Here is a snippet of the raw stack after DBE (note this is drizzled data but presented at native res, i.e. 1:2), autostretched:

Here is the stack after DBE and traditional deconvolution vs. the stack after DBE + BlurX in pixinsight, both autostretched.

To my eye the BlurX achieves very similar detail enhancement, but without the additional artefacting/noise generation that traditional deconvolution introduces. I can't see any 'fake' detail, though I'm sure you would get such results if you push the tool too hard. I personally prefer a light touch when processing in general anyway; it's very easy to push things too far with a lot of processes and tools.

I think this is an important comparison to make as it shows the difference from the traditional tools. I can say without a doubt that BlurX is hugely simpler to use compared to traditional deconvolution. The star distortion correction (pushing psfs towards circular) is also very powerful for those with less expensive optics.

I'll be trying it on a few other images but my first impression is certainly positive!

[silentrunning]

 

All examples straight out the box with no additional processing and default parameters.

 

Edited December 16, 2022 by silentrunning

[ONIKKINEN]

Its getting really difficult to resist getting PI based on the results so far. Looks like a TopazAI but built for AP from the ground up so fixes the "Topazed" look of that tool.

[ollypenrice]

I've had a first play this afternoon, working on a small galaxy, NGC772. BlurX has improved the image without doubt. It got me a sharper inner galaxy when compared with my original. However, my original had a painstaking sequence of Unsharp Mask iterations of different scales, selectively applied in layers and taking an hour or so to do. The BlurX image beat it after just a basic stretch. As Adam Block points out, the difference is even greater when the image is taken beyond 100%. The BlurX one isn't bad at 200%. I also think it got me deeper into some very faint tidal tails, but I can't be sure at this stage because the first process was a long time ago.

I looked hard for fundamental differences in small structures between the BlurX version and the Unsharp Masked version. There are none. The sharpening pulls out the same details but Blur X does it a tad better and with a tenth of the effort (and less noise.) I also made a careful comparison with a professional image of this galaxy and could find nothing to suggest the invention of spurious detail in mine.

I'll be buying it.

Olly

[geeklee]

Colour & noise aside. The BlurX one on the bottom has barely had anything done to it.  The one on the top took me many hours last year.  BlurX resampled to match original finished image.

Just cropped and AutoSTF for comparison.  Hmmm....

EDIT: Ran BlurX with reduced settings so it wasn't so glaringly different.

With BlurX settings quite high:

Edited December 16, 2022 by geeklee

[Bibabutzemann]

I was always scepitcal towards AI shapening because i felt like it invents new data but this approach seems different. 

I just wish there would be a standalone version, since i really like my workflow with SiriL and PS. I dont see myself paying 350€ to use it.

Edited December 16, 2022 by Bibabutzemann

[gorann]

I like what I have seen and tried so far and my experience with the other RC filters made me buy it before I tested it, and fortunately I am already using both PI and PS, although a PS version would be slightly more convenient for me. Since the other actions by Russ has been released for PS, it may only be a matter of time for a PS version.

[tooth_dr]

30 minutes ago, gorann said:

Since the other actions by Russ has been released for PS, it may only be a matter of time for a PS version.

Possibly, here’s the official response:

[gorann]

2 hours ago, ollypenrice said:

I've had a first play this afternoon, working on a small galaxy, NGC772. BlurX has improved the image without doubt. It got me a sharper inner galaxy when compared with my original. However, my original had a painstaking sequence of Unsharp Mask iterations of different scales, selectively applied in layers and taking an hour or so to do. The BlurX image beat it after just a basic stretch. As Adam Block points out, the difference is even greater when the image is taken beyond 100%. The BlurX one isn't bad at 200%. I also think it got me deeper into some very faint tidal tails, but I can't be sure at this stage because the first process was a long time ago.

I looked hard for fundamental differences in small structures between the BlurX version and the Unsharp Masked version. There are none. The sharpening pulls out the same details but Blur X does it a tad better and with a tenth of the effort (and less noise.) I also made a careful comparison with a professional image of this galaxy and could find nothing to suggest the invention of spurious detail in mine.

I'll be buying it.

Olly

Olly, just as we thought it would be more than enough work to re-do everything we had in stock with StarXT and NoiseXT, we now also have to do it allover again with BlurXT. Good there are cloudy nights!🤣

Göran

[Budgie1]

9 hours ago, Budgie1 said:

I bought BlurX last night and haven't received the activation code email yet. I did get one signing me up to the updates email list but nothing else. 

I emailed Russ and he resent my licence details, so BlurXTerminator is now installed in PI and working.

Here's my first comparison using it:

 

[ollypenrice]

OK, two results. This is NGC772 reprocessed. Note the full screen icon upper left and click on the image for the largest version. The BlurX images are shown at full size whereas I had to reduce the originals a little. These are all heavy crops from the TEC140/Atik 460.

https://ollypenrice.smugmug.com/Other/Galaxies/i-hH2SsDK/A

The previous version is below. Sorry about the change in orientation.

https://ollypenrice.smugmug.com/Other/Galaxies/i-vDD6Gw3/A

 

Now NGC891. I was always disappointed by this image because I just couldn't get the hairy little strands at right angles to the galactic plane to look right. Using unsharp masking to pull out the detail did a lot of collateral damage. One look at the data directly after BlurX showed that my problem was solved. This is the new version:

https://ollypenrice.smugmug.com/Other/Galaxies/i-SZLsvmF/A

The original is below.

https://ollypenrice.smugmug.com/Other/Galaxies/i-V36HW2z/A

In both cases the use of Blur X produced better detail with much less intervention, gave a cleaner result and, while the difference at 100% is significant, at 200% it is staggering. This may also have an impact on print quality. I haven't tried a print, yet.

Olly

 

 

Edited December 17, 2022 by ollypenrice
clarification

[ollypenrice]

5 minutes ago, ollypenrice said:

 

 

 

Sorry, false click.

Edited December 17, 2022 by ollypenrice

[Laurin Dave]

Excellent improvement Olly..  a quick trial suggests that the Fireworks (and also the Hidden, although I've never imaged that) galaxy would benefit greatly from Blur Ext treatment 

[teoria_del_big_bang]

What a great thread and discussion, I am off to at least get the trial version and give it a go 🙂 

Steve

[ollypenrice]

2 hours ago, Laurin Dave said:

Excellent improvement Olly..  a quick trial suggests that the Fireworks (and also the Hidden, although I've never imaged that) galaxy would benefit greatly from Blur Ext treatment 

I didn't see this because I was busy reprocessing -  yes - the Fireworks galaxy!  And you're right, the combination of StarX and Blur X (with a little help from Noise X) has made this into an altogether new prospect. The poor old galaxy was swamped by stars and had lots of little details which got knocked about by regular sharpening routines. I felt I was able, this time, to keep a softer look (my current undertaking) while finding more detail into the bargain. The Ha contribution is very discrete in order to preserve that gentler look but could easily be turned up. This is posted on my gallery at 100%. There are mild JPEG losses but I didn't want to downsample.

https://ollypenrice.smugmug.com/Other/Galaxies/i-dhnLzws/A

Olly

 

Edited December 17, 2022 by ollypenrice
clarification

[wimvb]

The game changer isn't just in what BXT does to the main target, but just as much in what it does to the background. With ordinary deconvolution, the background is never sharpened, because the noise gets in the way.

I just downloaded an evaluation copy and used it on my version of ngc 7331, where I had previously tried to enhance the ifn and satellite galaxies. This is a crop of a small background area. BXT pulls out background galaxies, that also show up, but are unidentified in Aladin (SDSS)

(UG = Unidentified Galaxy)

Here's the same area from Aladin