Processing in AI (split-off thread)

[jaapv]

Advertisement (gone after registration)

Maybe time to split the thread for clarity. 

[pgh]

3 hours ago, jaapv said:

I try to define the point at which we draw the line

I’ve been in discussions about this with my colleagues for years. (Actually, now I’m old enough to say decades, unfortunately).

Here’s what I’ve concluded.

No line can be drawn. 

What matters for me is intent, and the context that the work is presented in. Sound arguments can be made for various methods to be used in conjunction with various intents in various contexts. I won’t opine on what one should and shouldn’t do.

I speak to what I am interested in - which is some sort of mission of fidelity to the visual record of a photograph (which yes, is already a record distorted by optics, flawed in all kinds of ways), and where that can get you. A photograph does not represent the world in any sort of complete or objective way, but it is also not disconnected from it. What is interesting about the world as depicted through photography is what is imperfect about it, primarily. 

Tools that try to push an image towards some arbitrarily defined “perfection” (i.e. de noise) have never been useful to me because I’ve never seen an image of my own be made better by it, nor anyone else’s really. But I don’t see grain, or noise, as an imperfection, for the most part. Yes, there are times it overwhelms a picture to where the legibility of it is undermined. I’ve not seen really examples where de noise has helped that - those pictures are too far gone in my experience. But that’s subjective. Maybe I’ve not seen the right pictures. Maybe others feel it does really help bring out the real content more clearly. All good, not going to argue about that. But cloning out elements (definitely not a universally accepted general practice, by the way) is the same thing - the Kent State shooting picture, notorious for being an earlier example of such selective removal, is not a better or more interesting photograph with the pole removed. 

Part of this for me is that using AI to “perfect” pictures usually seems to take an image that isn’t worth the work in the first place and just makes it more soul crushingly boring. If the picture is a strong one, it will be strong with the noise still, and not stronger without it.

YMMV. 

[chris_tribble]

If you’re in London and haven’t seen it do get to the Dennis Morris exhibition at the Photographers Gallery (https://thephotographersgallery.org.uk/whats-on/dennis-morris-music-life

I don’t think that Denoise would have enhanced his work!

[SrMi]

12 hours ago, chris_tribble said:

If you’re in London and haven’t seen it do get to the Dennis Morris exhibition at the Photographers Gallery (https://thephotographersgallery.org.uk/whats-on/dennis-morris-music-life

I don’t think that Denoise would have enhanced his work!

You cannot compare the aesthetically pleasing noise of the film with the ugly noise of digital files. There are even tools to simulate film noise and add it to digital files.

[LocalHero1953]

35 minutes ago, SrMi said:

You cannot compare the aesthetically pleasing noise of the film with the ugly noise of digital files. There are even tools to simulate film noise and add it to digital files.

Subjective call.
Digital noise varies widely between cameras. I found the Q2 noise to be ugly but the SL2-S noise acceptable.
Personally there is a lot of film noise I find distracting and would like to remove - sadly digital denoise tools don't work well on scanned film.

[LocalHero1953]

I should have added: the lesson I would draw from the Dennis Morris exhibition is not that his photos were enhanced by noise, but that noise is irrelevant compared to what his photographs actually show.