Is anyone using Topaz Photo AI for sharpening/denoise? I am trying to decide whether to buy it - it is quite expensive at $199. Is it worth it?
Also, do people consider it cheating since it uses an AI algorithm? I noticed that competitions such as Astronomy Photographer of the Year require you to declare the use of Topaz and other AI assisted processing tools.
I've really been trying to level up my processing and I wonder if Topaz will help. Here is my latest process from footage taken last year with my Lunt LS60MT. I would welcome any feedback on the processing I've done here. I'm never sure if I have the sharpening right and worry about overcooking.
Warren, here's a positive version from when I originally took the shot, my processing has come along a bit since then. For the inverted image above I did the inversion during the first step in IMPPG so I don't have a positive fully processed version.
Interesting topic.
Im using AI for upscaling movies, games, etc., but when it comes to solar imaging im oldschool and dont use it.
On social media i see a lot of solar images that excessively use AI alghorithms to get rid of noise and / or sharpening.
I see this as a sad developement since many of the interpolated / generated structures seen in these images are surely astonishing in a way that they impress masses of people that have not done solar imaging before, but these structures are simply not real. The line between a bit of denoising and generating fake structures is very thin.
I guess there will be a real boom of people using these tools in astrofotography until a point where they realize that this is more art then actually showing nature as it is.
Stacking a bit more images will get rid of any noise / add real signal to your image, so there is really no need to use these tools for this, especially in solar imaging.
I guess in the end it comes down to our intention, when we create solar images with a telescope.
When i use a clumsy and heavy setup with a 235mm aperture and high transmission filters my intention is to collect some real data.
I usually use Photoshop to reduce noise. It probably is not the most effective and maybe I should use another option. It a denoise program designed for solar available for MacOS?
James
Lunt 8x32 SUNoculars
Orion 70mm Solar Telescope
Celestron AstroMaster Alt/Az Mount
Meade Coronado SolarMax II 60 DS
Meade Coronado SolarMax II 90 DS
Meade Coronado AZS Alt/Az Mount
Astro-Tech AT72EDII with Altair solar wedge
Celestron NexStar 102GT with Altair solar wedge
Losmandy AZ8 Alt/Az Mount
Sky-Watcher AZGTI Alt-Az GoTo mount
Cameras: ZWO ASI178MM, PGR Grasshopper, PGR Flea
Lunt, Coronado, TeleVue, Orion and Meade eyepieces
Dennis, I share some similar feelings about not wanting to add artificial data. I do believe that processing is an artform and plenty of flexibility in how to go about it is a good thing, but I don't wish to add structure that wasn't originally there. For now I am just using Autostakkert, IMPPG, and Photoshop for my workflow. I think I am going through a process of deciding how many additional tools, if any, I want to add, and where I want to draw the line.
James, I don't think Topaz is specifically designed for solar but it does work on Mac (I like to use Mac and do my capture and Photoshop work on Mac, with Autostakkert and IMPPG on a Windows machine as an in-between step).
Just to come in from a tangent here and play devils advocate:
Why do you need to denoise an image in the first place? My question would be where is the source of the noise? If it is stacking then more frames are needed, if it is in post processing then the source of noise is oversharpening either through too small kernel size or excessive use of sharpening iterations. Both scenarios removing noise at source has simple solutions...
Just my 2 cents...
http://brierleyhillsolar.blogspot.co.uk/ Solar images, a collection of all the most up to date live solar data on the web, imaging & processing tutorials - please take a look!
Enhancing real image data is one thing. Creating false (fake) data using AI is quite something else, no matter how well intentioned. I’d say we should aim to keep our images pure and free from AI artefacts. Use any tools available in a way that’s sensitive and respectful to the purity of our craft. We can, and should, tweak every last drop of image data from our dataset, but no more, as then we’d enter the realm of fantasy.
Just my thoughts…
Stu.
H-alpha, WL and Ca II K imaging kit for various image scales.
Fluxgate Magnetometers (1s and 150s Cadence).
Radio meteor detector.
More images at http://www.flickr.com/photos/solarcarbon60/
Hal as a fairly reasonable Solar imager, what ever process you used on the first image, I would not change as it on the money. Everything can of course can be tweaked, nothing wrong with that. Topaz AI, I think, has no place in Solar processing. Same goes for planetary. It certainly has a place but not for what we do.
Ralph.
Vixen 130ss SolarMax 90 BF30 EQ6 Pro, LX200 14, C8 CG5, Megrez80, Lunt 100 DS. EQ 8
ASI 120mm 290mm 432mm Lunt B1800 Cal-K + other stuff.
I don’t see any real artifacts in the image from the processing you have done. In the past I have seen different folks post some images (on a different site) where they claim they came up with a process to bring out details that were hidden by hazy conditions and I knew they were using some presets in Topaz that are strictly for artistic effect. Mark is spot on that much of the noise can be caused by over sharpening when trying to bring out details that are either beyond the seeing conditions or the capability of a particular set up. I like to remind myself that I’m imaging hot plasma and gas and that simply will not have tremendously sharp features compared to the moon which is basically a rock!
I have no issues with folks using software like Topaz to create something artistic with their solar images as long as they outline what they did and not try to pass it off as something no one else could possibly capture! Like many “news” sites might try with the announcement that “Jupiter will be the size of the full moon tonight. Don’t miss this rare occurrence!”
Warren
Lunt LS60T DS
Orion ED80T CF
Meade ETX LS6
Lunt CaK BF1200
Lunt WL Wedge
Baader Photographic Film
ASI174MM
Skyris 236M
Player One Saturn-M SQR
I use topaz denoise for all of my wildlife photography and have found nothing that comes even close to the same quality.
I have photo AI, Denoise, sharpen, and every other product they sell and I have found that they do not work as well on monochrome solar images as RGB 12-16 bit images.
As far as using AI goes, I treat it like any other tool. As long as the end result is a clean and realistic approximation of what you actually see, I have no problem using new tools.
It’s when these yahoos get on Instagram and act like they’ve created the first solar image ever made and go on and on about all their processing techniques, etc. and then post a prominence animation with one of those AI motion programs when I think they have gone too far.
when we all started this back in the early 2000s there were about a dozen people on earth doing this hobby and we were discovering all the new ways to process and capture images, etc. we may have created a monster by popularizing the hobby so much because every day somebody new is putting up a solar image processed using the techniques that we discovered and then claiming that they invented it and this is the “greatest solar image ever ever made”.
This guy Andrew McCarthy on instagram cracks me up with his self aggrandizing.
Stephen W. Ramsden
Atlanta, GA USA
Founder/Director Charlie Bates Solar Astronomy Project http://www.solarastronomy.org
Thank you for your comments and discussion everyone. It has been very useful and has validated some of my feelings of reticence about using AI. I think I have decided to stick with my current software and work on improving my workflow without adding AI tools.
Stephen - I started with the transit of Venus in 2004 when I was at university - 20 years ago! Baader film and a little Meade telescope with a compact digital camera held up to the eyepiece. My imaging has come on a bit since then!
marktownley wrote: ↑Sat Mar 02, 2024 7:59 am
ust to come in from a tangent here and play devils advocate:
Why do you need to denoise an image in the first place? My question would be where is the source of the noise? If it is stacking then more frames are needed, if it is in post processing then the source of noise is oversharpening either through too small kernel size or excessive use of sharpening iterations. Both scenarios removing noise at source has simple solutions...
Just my 2 cents...
Very good point Mark !
I've never used denoise in solar imaging or in lunar imaging. If denoise is needed, it means there is something wrong somewhere (acquisition or processing).
I use a gentle touch of denoise in planetary imaging, and a bit more for very difficult targets such as Uranus or Neptune.
solarchat wrote: ↑Sat Mar 02, 2024 6:50 pm
This guy Andrew McCarthy on instagram cracks me up with his self aggrandizing.
I think he likes to use Electric Edges in his images from Topaz Studio.
Normal Ha Disk
FD-2-X-Barlow-Ha-2-18-2024-Normal.jpg
Electric Edges from Topaz Studio
FD-2-X-Barlow-Ha-2-18-2024-Electric-Edges.jpg
I very much prefer this without "electric edges"! And look at all those filaments! <3
Agreed! That is my image, here is one of his and you should see why I think Electric Edges was employed. You just don't see sharp lines quite like that.
mccarthy-3.jpg (663.83 KiB) Viewed 540 times
Warren
Lunt LS60T DS
Orion ED80T CF
Meade ETX LS6
Lunt CaK BF1200
Lunt WL Wedge
Baader Photographic Film
ASI174MM
Skyris 236M
Player One Saturn-M SQR
nothing on the sun is sharp and there are no sharp edges.
We get people like McCarthy show up every couple years and then just keep pushing it further and further for more social media attention until they get caught doing something really stupid and fake.
Anyway, to each his own.
Stephen W. Ramsden
Atlanta, GA USA
Founder/Director Charlie Bates Solar Astronomy Project http://www.solarastronomy.org
Being told that my early images were over-sharpened might have put me off continuing.
Instead I took it on board and gratefully accepted the advice as valid and vital to personal imaging progress.
The simplest way to avoid over-sharpening is to avoid completely artificial, uniform texture.
Does your image look as if it is projected onto coarse cloth or even sacking? Or had a terrifying attack of the AIs?
Above all: Compare your results with everybody else's. Do your images look like theirs?
There are regulars here who can pull off remarkable images through cloud and awful seeing.
Their repeated success, under the most trying circumstances, is due entirely to their well practised skills at capture and processing.
NOT by over-sharpening by all artificial means. Comparisons are NOT odious when it comes to solar imaging.
They are the fastest way to making real, personal progress. So that you record the nearest thing to humble reality.
The AI images posted here were absolutely hilarious. IMO.
Does the "artist" actually own any solar equipment?
Or is it entirely text prompt to AI image?
DanielBrousseau wrote: ↑Mon Mar 11, 2024 4:18 pm
I used Topaz AI on my best shot (version 2 first here) and tried Topaz. It's clean and a bit artistic for me, is it too much for the purists ?
Idk, looks a bit oversharpened to me. I applied some unsharp masking to the first one and i wouldnt push it further. Its good material, but quickly oversharpened.
Only my viewpoint.
Attachments
AR3435_22 sept 2023_version 3.png (1.64 MiB) Viewed 347 times
Lunt LS50THa, B600, single stack
William Optics Z61, Baader solar film filter
DayStar Quark Chromosphere, w/1.25" ZWO UV/IR filter
William Optics Z81
ZWO ASI178MM
DanielBrousseau wrote: ↑Mon Mar 11, 2024 4:18 pm
I used Topaz AI on my best shot (version 2 first here) and tried Topaz. It's clean and a bit artistic for me, is it too much for the purists ?
