Output Transform Tone Scale

Discussions - ACESNext VWG – Output Transforms
244

Could we get a rational for this?
ARRI ALF, ARRI REVEAL, RED IPP2, TCAM all raise mid grey a bit when going to higher peak luminances.
ACES 1.0 had it fixed and people generally didn’t like it. Hence the SSTS introduced a similar grey boost.

The suggested tone scale very much models natural system.
A natural system does not pin a single value or part while others change…

Maybe it would make sense to make a larger survey before we bolt this requirement in.

You need to verify what subsequent processes are driven by the rate of change of the tone scale, and if you want to give up continuity and simplicity.

245

I will get them to ring in.

It’s not about it being the default, but rather allowing them mid gray output luminance values on various dynamic range displays to be matched.

246

What should be matched?

247

The SSTS could match mid gray (and the rest of the tone scale) via an exposure adjustment. The SSTS achieved brighter mid-gray but shifting the entire tone scale to the left, effectively moving the mid-gray to a higher output luminance. This mimics what happens when you make a print in a negative/print system.

248

edited the original post for clarity

249

We raise grey also by an exposure adjustment now? I do not follow, I am afraid.

250

I think @sdyer has plots that clarify

251

And have we agreed that ACES 2.0 is a parametrised transform? How do we track those parameters? What about the archives? Do they change per shot?

252

Also an absolute luminance match is not a perceptual match, right?

253

Parameterized is what sense? What do want the parameters to control?

I would suggest it’s a bad idea to have the output transform change per shot.

Instead I’d suggest that there’s a single output transform algorithm with parameters that adjust it for a given display’s characteristics (e.g. max Y, min Y, primaries, etc.) I don’t think these parameters should be used for creative control. That’s what grading is for.

254

Agreed … I think they literally want a match on the meter for mid-gray

255

Good we are on the same page.

Then we need to choose, either we let grey be brighter with brighter peak luminance or we pin grey at the same luminance. We cannot have both.

If we agree to make grey brighter with a greater peak luminance than the suggested curve works fine.

256

Then we need to explain them, that this might be not a good idea…

257

I agree with you.

My comments during yesterday’s meeting were premature. I’ve been tasked with looking at this and I think our current curve has exactly the behavior we want.

1 Like
258

Ok, glad to be on the same page on this one…

1 Like
259

Thanks! What would be the rational for a fixed mid gray?

It is like putting your knee in the back of the curve and trying to stretch it by the should to handle higher peak luminance, something is going to yield…

260

I’ve made a lot of SDR/HDR comparisons in the past days and weeks. I’ve noticed that the 100 nits curve appears brighter in the shadows compared to the 1000 nits curve. My guess is that the 1000 nits curve just appears darker / more contrasty due to the brighter highlights.

Maybe the 1000 nits curve needs less flare compensation / shadow toe? Can anybody confirm my observation?

1 Like
261

I agreee. The flare for >100 nit curves is driven by how much the middle gray is shifted higher. I’ve argued for lowering it which, in my testing, has resulted to better appearance match between 100 nits and 1000 nits (contrast matches better). I’ve tested by lowering the w_g parameter value to 0.1. The original value was not derived by looking at images but by looking at curves.

I’m hesitant changing the toe parameter itself as that would change the 100 nit curve as well.

262

I agreed the flare parameter is quite high. I think the originally suggested setting was lower.

1 Like
263

I’ve done some additional tests today. With w_g increased to at least 0.3 the problem disappears. I’ve also analyzed the HDR export in Adobe Lightroom and the SDR conversion in Adobe Photoshop. They are equivalent to w_g ~ 0.4.

Besides, I find it strange that I always have to increase the contrast significantly in a node before the DRT to get a good, yet neutral looking image when using my own camera footage. (t_1 reduced to 0.0 in this case.) Are the ACES sample frames really neutral in terms of contrast?

1 Like