Blur exterminator

[azure ore]

Have you guys seen the blur exterminator plug-in for pixinsight? It does such a good job!

[jovial root]

would you wanna give my orion a try? does it take PNGS?

[hybrid meteor]

Yeah blurX is cool but sometimes can actually create fake structure that aren’t in the image and if you spend time learning Decon you can do better than it without creating fake structures

[stone hull]

could you do mine too?

hold on

[jovial root]

what the hell happened with your image

[sonic lion]

also blurx wont help that, you need a priest

[stone hull]

lmao

internal reflection but idk

its been cloudy for the last month and a half so i cant re shoot

[jovial root]

you can make a new layer that only has the reflection using a dust and scratches filter and subtract the colors from the original image

if you have photoshop

[low token]

The AI used in bXt was trained on actual space images that have deep detail resolved, which enables bXt to understand what details are there, so it's not necessarily "fake" detail, but I see your point. Also, I've seen Adam Block's video on bXt and performing decon manually, while getting results anywhere close to bXt, is quite difficult. But all in all it's personal preference.

[hybrid meteor]

Yeah… So it when set too strong will create a structure that isn’t there in your image but in the based image (or just completely fake) and vikram explained it very well how it works and why it happens and confirmed it does happen. I use manual Decon and it takes about 5 minutes to beat blurX saying blurX isn’t set to high

[low token]

I'll have to look into manual decon more, I'd actually rather get similiar results for free than to pay for bXt.

[desert flower]

uhhh

im not entirely sure about this statement

Deconvolution is just a tool blurx uses

Blurx is effectively an easy way to deconvolve your image

if you manually tweak your psf, you can get more realistic/better results

but manual decon will never achieve blurx

the best processers here cant get to it

[hybrid meteor]

Wrong? Me and zegery have came up with something super simple that works very well and depending on data beats blurX but Vikram did a long paragraph awhile ago explaining why blurX is good and why it can create fake structures I can’t remember the specifics of it but he was very in-depth with it but it was awhile ago when it was brand new

[desert flower]

Im aware of what vikram said

if you can show results where manual deconvolution beats blurx id be very surprised

but please show

[hybrid meteor]

Sure give me 2 seconds

Top or bottom?

This one was a little noisy as we was working on it but

[desert flower]

but this data is processed differently

[hybrid meteor]

No?

[desert flower]

in the case of the head on

or whatever that is

its definitely processed differently

the colours and contrast levels are different

[hybrid meteor]

I duplicated it in pix and used one with manual and one with blurX

So its definitely the same

[desert flower]

uhuh

the second one is also clipped

not to mention some weird denoising going on on m51

[hybrid meteor]

Again different decon methods will change the image differently and yeah M51 was a practice run when we was first testing

[desert flower]

which ones of the two are manual and blurx

[hybrid meteor]

Bottom for both im pretty sure

[desert flower]

?

bottom for what

[hybrid meteor]

M51 and NGC 891

Bottom image

of the two

[desert flower]

yes what did you do to that

I asked whether they were manual or blurx

[hybrid meteor]

Yeah the bottom images are manual and the top ones are blurX

[desert flower]

very bold claim you're making then

[hybrid meteor]

I know im very aware

As i said it really depends on data

[desert flower]

The conclusion is that no, manual deconvolution cannot recreate blurx...

You've shown two examples where it's worse though

so

thats not really

you cant even conclude that it depends on the data

[hybrid meteor]

I mean ive asked Hermit dylan which one was better before without him knowing which is which he picked ours

idk what to tell you

[desert flower]

the core speaks for itself....

more noise + less detail + more clipped

[hybrid meteor]

Yeah but again its progress. BlurX is a strong tool but then i look at Stefans decon on the BATs stuff and the details in his are so much better than BlurX hwich is why i dont doubt its possible

Our test is literally like a 5m decon that doesnt take long. Where BlurX is a decon that would usually take hours to get to which considering a five min decon it really isnt that bad, i know about the noise but since then we have worked on fixing the noise which i havent got examples of rn but its getting rid of it so far. For the clipping i believe thats just the mask i used to not effect the background too much

[toxic patrol]

that can be fixed

[desert flower]

the details look worse

like im not gonna put it nicely, the manual deconvolution is worse

not by a marginal factor

[toxic patrol]

for that image i agree

depends how you use the decon

[desert flower]

ofc

blurx just so happens to use it better

[toxic patrol]

the ai is very good with deconvolution

with enough practice it can be beaten

[toxic patrol]

this image is a bad example of it being 'beaten'

[desert flower]

agree to disagree

[toxic patrol]

background grain aside which could be solved with masking I would say the manual came out nicer

[desert flower]

this is because the top one hasnt been denoised

its bad practice, but with noisy data denoise -> decon yields the best results

I am saying this from experience with m51, my data is very noisy, and by denoising it i've managed to pull crazy detail

[toxic patrol]

Understood

For the comparison there was no other major editing other than deconvolution

For another test we would have to process them better before and beyond that state

That way its fair to blurx

[hot raven]

I need cloudxterminator

[latent stag]

I don't have access to this, Could someone try it on either of my images please? I think they're a bit blurry.