Quote:
Originally Posted by Mickoid
I know where you're coming from Marc and I agree - partly. I guess I'm also thinking of shots using the Hubble palette, starless views of nebulae, planetary shots with their overexposed moons combined in the scene. They all look pleasing but are they representative of what came through the OTA to the sensor? Photography is an art and can be a form of science as well. I personally like to keep things as natural looking as I can but also fall into the trap of over sharpening or over saturating my shots to make them look pleasing. It's a fine line and defining 'real' is difficult, defining 'art' is conjectural.
|
All good questions.
The HST palette is an arbitrary assignment of line emission data to color channels or blend when you state you're doing a 50/50 bicolor, usually the green channel being the data mix, the other two red and blue being the full data respectively. There is nothing wrong in doing that. If you further mask and change the color balance and start mixing color channels to suit then it enters the realm of art. Because you're no longer representing the accurate distribution of gas in the nebulae and the physical processes that are at work. But if that floats your boat then go for it as long as you're upfront about what you're doing and are aware it doesn't really mean anything anymore. It's become your personal rendition and it is no longer rooted in science.
My understanding is that as soon as you stretch data and it's non linear it has no scientific value. We all stretch our dynamic range trying to show as much as we can without it looking too stupid and HDR it all and "respecting the light". So the balance is subjective and already an interpretation of what the real data is for the sake of displaying it. We're not changing the data. Only stretching it.
Then comes the noise reduction, sharpening and possibly star removal. That can create processing artefacts if you overdo it. If you do it manually or with the appropriate software then it's called processing the data. If you use software such as Starnet++, DenoiseAI, SharpenAI, then you let the software introduce new data that didn't exist in your original signal based on what it was trained on. Mostly terrestrial shots. Again if that floats your boat go for it. But be aware it's fake and be upfront.
Now ask yourself why do you spend hours outside at night capturing photons if you're going to discard them in the post processing stage and let a piece of software substitute them so it looks cool? Sounds a bit counter-intuitive. I think people who use TopazAI like what they see. It gives them an edge because most don't disclose it. It's becoming harder to see what's real or not. You just have to look at Cloudy Nights, but mostly Astrobin where even some people judging shots are "AI evangelists".
I like looking at a shot not having to wonder if it's legit.