BlurXTerminator - Wow!

[teoria_del_big_bang]

I am off for a drink or two with some old school mates, some I have not seen for 45+ years so will not be able to comment, as I cannot see well enough on my mobile,  but please let the thread continue to be a good and informative discussion, I would hate to come back to it and see it closed down or have a telling off 🙂 
 

Steve

Edited December 29, 2022 by teoria_del_big_bang

[newbie alert]

38 minutes ago, Stuart1971 said:

Anybody that can’t get there head around PI should get another hobby, as the rest is much harder….IMHO…

I've never used it, but I can't get my head around why there's so many stupid names, is it to make it sound technical , to make a simple task sound complicated... 

[ollypenrice]

We managed a first light with our Samyang 135 last night. Most of the stars are very good and all corners hold up for viewing at 50% of full size. At 66% they look acceptable to me but I am absolutely not a pixel peeper and don't want to become one.  At full size two corners show clearly distorted stars.  However, I ran the image through BlurX because some users report improvements in distorted stars. In my case it made them worse on a first attempt but I need to try different parameters. I'll update this finding as I experiment. Note, I was working with a single sub and Blur X does require a good S/N ratio to give its best.

Olly

[ollypenrice]

3 minutes ago, newbie alert said:

I've never used it, but I can't get my head around why there's so many stupid names, is it to make it sound technical , to make a simple task sound complicated... 

I don't think they realize what their language sounds like to most people. It's a progam written by geniuses for geniuses - so that's me out of the frame!

Olly

Edited December 28, 2022 by ollypenrice

[newbie alert]

From viewing this thread it seems like a useful tool but since its release plenty of images to prove that people are mis using it to over sharpen, showing artifacts from the Ai technology that aren't really there within the data captured.. sticks out like a sore thumb... 

[newbie alert]

5 minutes ago, ollypenrice said:

I don't think they realize what their language sounds like to most people. It's a progam written by geniuses for geniuses - so that's me out of the frame!

Olly

I must be well down that ladder then

[geeklee]

9 minutes ago, newbie alert said:

I've never used it, but I can't get my head around why there's so many stupid names, is it to make it sound technical , to make a simple task sound complicated... 

Like , , ,  and  ?  

Why are they "stupid" names?  Aren't they exact descriptions of what the process is doing?  

You can drag the Process to your Workspace and rename it whatever you like.  I believe MT can be used for other things of course.

 

[scotty38]

18 minutes ago, newbie alert said:

I've never used it, but I can't get my head around why there's so many stupid names, is it to make it sound technical , to make a simple task sound complicated... 

Come on  please, there's enough unnecessary jumping on the "let's bash pixinsight" train as it is without admitting it's never been used but it's still rubbish 😀

[ollypenrice]

The PI thing should be another thread. Let's stick to Blur X.

Olly

[geeklee]

28 minutes ago, ollypenrice said:

I'll update this finding as I experiment. Note, I was working with a single sub and Blur X does require a good S/N ratio to give its best.

I'd be interested how you get on with a stack as I found a slight improvement - if that - but nothing like some other non Samyang images I've seen.

[DaveS]

Well, I don't know about "written by monkeys" but it's certainly written by mathematicians, for mathematicians, possibly working in a little hermetically sealed bubble, with no reference to the outside world.

I did start the 45 day trial, even bought Warren Keller's book, but it still did my head in.

That being said, I've been looking at some of Adam Block's YT videos which make things a bit clearer, so will likely be purchasing a license in the new year, when I have a new computer with the power to run the software well.

[geeklee]

<Error posting>

Edited December 28, 2022 by geeklee

[newbie alert]

23 minutes ago, geeklee said:

Like , , ,  and  ?  

Why are they "stupid" names?  Aren't they exact descriptions of what the process is doing?  

You can drag the Process to your Workspace and rename it whatever you like.  I believe MT can be used for other things of course.

 

Ok histogram transformation... Levels

Curves transformation... Curves( pull it up or down)

Channel combination.. merge channels

Why do you need a script to do this.. it's just a click

[Magnum]

im at 50 mins into the video with M1 the Crab as the target. I think that is an excellent choice to of target to experiment on, as it's the Object that always shows sharpening artefacts the most.  Adam is right with his description, this type of sharpening usually tends to connect structures making them worm like or with M1 I most often see a honeycomb artefact introduced.

Since Topaz AI became popular I see these artefacts all the time on on posted images online and it really really annoys me, I sticks out like a sore thumb but amazingly those tend to be the images that get the most WOW comments and likes ( go figure ).

I like that BlurX can be controlled to prevent this with the sliders ( though if im sharpening in photoshop the sliders work in real time rather than having to apply it see the result, instead of trial and error with different slider positions with repeated apply and undo until happy with the result.) This is my biggest frustration in Pix and Topaz AI, that real time previews are only available in some processes as far as I can tell from my limited use of the program. I usually have to take the result from each process back to PS as a layer on top of the previous version then adjust the opacity until I think it looks natural but not overdone. Or selectively rub through only the bits I think should look sharper. 

what would be nice if there was a way to prevent people from going over the top with it, as its so easy to get carried away when sharpening and thinking its getting better and better, but there is a point where it just starts looking like a water colour or digitised image, being AI there should be a way to make an intelligent limit to prevent over sharpening, based on the structures joining together. 

Lee

Edited December 28, 2022 by Magnum

[newbie alert]

25 minutes ago, scotty38 said:

Come on  please, there's enough unnecessary jumping on the "let's bash pixinsight" train as it is without admitting it's never been used but it's still rubbish 😀

When did I say it was rubbish?? 

Don't jump on things that wasn't there in the first place

[scotty38]

2 minutes ago, newbie alert said:

When did I say it was rubbish?? 

Don't jump on things that wasn't there in the first place

I must admit I thought you would pick up on my poor choice of words but my fundamental point stands.

[ollypenrice]

1 hour ago, Magnum said:

 

what would be nice if there was a way to prevent people from going over the top with it,

Lee

I think this is true with most new processing tools. In the early days of Pixinsight, people went absolutely crazy over its HDR wavelets, prompting Dennis Isaacs to ask, 'Why do all Pixinsight images look like pictures of human brains?' It's true, they did, but folks have eased off with the tool and now use it invisibly most of the time.

 

I still can't get BlurX to improve my Samyang image.

Olly

[Magnum]

9 minutes ago, ollypenrice said:

 

I still can't get BlurX to improve my Samyang image.

Olly

I use mine stopped down to f2.4 then it’s almost perfect to start with. 

[ollypenrice]

49 minutes ago, Magnum said:

I use mine stopped down to f2.4 then it’s almost perfect to start with. 

An option, but that's a very significant loss in light grasp even though 0.4 doesn't sound like a lot. 4 hours becomes 5.76 hours just in terms of photon count. Do you stop down with the diaphragm or with an aperture mask?

Olly

[Magnum]

30 minutes ago, ollypenrice said:

An option, but that's a very significant loss in light grasp even though 0.4 doesn't sound like a lot. 4 hours becomes 5.76 hours just in terms of photon count. Do you stop down with the diaphragm or with an aperture mask?

Olly

I use the built in diaphragm as I really like the super fine 18 point spikes this lens produces, was one of the reasons I bought it. I know I’m in the minority with liking spikes, I don’t like newt spikes or any other type, only 18 points. And it only produced them on the very brightest stars.

lee

[The Lazy Astronomer]

This thread needs more images!! (and a bit less arguing)

If anyone's got a few comparisons of their data with no decon vs blurx vs the best they could do with manual decon (and no other processing) I'd very much like to see it. 

I don't necessarily mind doing deconvolution (it's somewhat cathartic, I think - playing around with the settings to try and get it right), but I never feel I've gotten the best out of it, purely due to lack of experience, so BlurX is a tempting prospect. 

[ollypenrice]

6 hours ago, The Lazy Astronomer said:

This thread needs more images!! (and a bit less arguing)

 

No it doesn't!  

Oh all right, maybe it does.    I  can offer these. They are heavy crops of what is, for a TEC 140 (1 metre FL) at 0.9"PP, a small target. The first is the original, sharpened in Ps using iterations of unsharp masking on different layers.

The second was sharpened first with BlurXT and then with a single iteration of unsharp mask. It has also been through StarXterminator for star reduction, so bear that in mind.

My feeling is that the second has more delicacy and finesse. In the first, the main dust lane has more impact but that's because it has been artificially widened and darkened by USM. The BXT image has the dust lane in better agreement with R Jay GaBay's large telescope rendition here. https://www.cosmotography.com/images/wide_ngc891_2010.html

My image had 7 hours luminance and 2 hours per colour, so 13 hours. That's a few hours shy of what I used to aim for doing galaxies in the TEC.

Olly

GGC891original.tif

[PeterCPC]

I have been playing around with this on some old images. As an example I have posted crops of image of the Rosette from this year without and with BlurX. There is an improvement but it's quite subtle I feel.

Will I be buying this? I don't think so. If it was a plug in for PS/AF then I might be more tempted but the price of £83 seems quite high for what you get.

 

 

Edited December 29, 2022 by PeterCPC

[ollypenrice]

In case you haven't watched one of the videos in this thread, there's a PI script available here: https://www.skypixels.at/pixinsight_scripts.html

Once installed (it appears in Script - Render - PSFImage) it will give you a PSF for your particular image, allowing you to over-ride the auto PSF on BlurXT. Some commentators are suggesting that it out-performs the present auto PSF. I think maybe it did on the image I'm exploring at the moment, our single sub Samyang first light. My feeling is that you can't really expect it to work on a single sub, though it worked a bit better when the sub had been noise reduced.

Olly

[scotty38]

I can’t recall the reasoning but I seem to think it’s advised to not do any noise reduction before using blurxt