ISO-based denoise presets useless for iso-invariant cameras

[rvietor]

@paperdigits

Then its simply matched from the ISO metadata.

But his use case wants a combination of “exif iso” and dt exposure correction. E.g. “ISO 100 + 4EV” should match to the noise profile for ISO 1600. Best bet might be a LUA script?

[paperdigits (Mica)]

@rvietor

But his use case wants a combination of “exif iso” and dt exposure correction. E.g. “ISO 100 + 4EV” should match to the noise profile for ISO 1600. Best bet might be a LUA script?

I don’t know, did profiled denoise at 1600 produce good results? I didn’t get that sense, but maybe I’ve misremembered or misread.

If PD produced good results at 1600 for this use case then sure, lua or a style/preset would probably work fine.

[paperdigits (Mica)]

@rvietor

I mean, you don’t need to edit all images you took up to presentation standard (let alone ready for printing).

We should also keep in mind that OP is likely satisfied with their workflow mostly.

[grubernd]

@paperdigits

This seems unnecessarily harsh and I’d request some empahty in future replies.

Yes Sir, duly noted.

@rvietor

As you wish

I am sorry, my answer was a bit over the top. In internet discussions a lot of sub-text can get lost.

@paperdigits

We should also keep in mind that OP is likely satisfied with their workflow mostly.

It’s less satisfied and more “that’s just the way it is”.

I like to have the edits I send out into the world to look decent. And a lot of reportage and event coverage has tons of images that are not among the selects but are needed as proof for advertisers etc. I don’t aim for perfect but more for consistent.

My discovery of an almost iso-invariant camera together with darktable scene-referred changed a lot how I look at exposure during taking pictures and editing. Of course it has a downside, hence this thread.

From what I learned these days, it should be possible to create some good-enough styles. Also it might be the thing that gets me into coding some Lua for darktable that does the required calculations for me.

[paperdigits (Mica)]

Thank you for your understanding! Much appreciated

[rawfiner]

In the ideal world, we would have perfect profiles, and in that case, the result of denoising would be as good whatever exposure compensation is going on afterwards. Unfortunately, the model does not perfectly fits reality, so indeed you have a result which is less good.

Though, you can play with the “adjust autoset parameters” to improve the result (usually, multiplying its value by 2 for each stop works quite well), or with “preserve shadows” (by lowering it. A value of 1 can give a tradeoff probably usable on most images, even though on darker images lower values work better).

Also, wavelets is much less sensitive to the model inaccuracy than non-local means, as such the easiest workflow in my opinion would be to use wavelets, with preserve shadows set to a constant value between 0.9 and 1.0.