Viewing environment compensation, and mid grey alignment, in ODTs

Discussions - ACES Development VWG – Output Transforms
1

Good Friday everybody,

Now, long after the ODTs of ACES 2.0 have been published, I have thought about the contrast-perception et cetera in our few reference viewing environments:

  • Question 1: Could, as an example, Rochester Institute of Technology, via a kind proposal from AMPAS and the ACES community, perform a study - at least regarding, but not limited to, the perceptual impact of our two, in practice most used, reference viewing environments?

The two rooms:

Room 1. UHDTV-, HDTV-, HDR-, SDR-grading irrespective of picture resolution (as many/most colourists use one single room for all their work): Viewing distance 3.2 picture heights; 5 nits (D65) of surround luminance; “zero” nits of periphery luminance; “zero” ambient lighting. (Based on Table 3 in ITU-R BT.2100-3).

Room 2. Cinema-grading: Based on SMPTE RP 431-2:2011, D-Cinema Quality — Reference Projector and Environment. Note: “The reference viewing position for colour grading shall be at a viewing distance of 1.5 to 3.5 screen heights …”, i.e., the impact of the viewing angle is likely different in comparison to Room 1.

The coarse surround ratios:

I believe that the current definitions by CIE of Average-, Dim- and Dark- surround ratios, (surround luminance / displayed luminance of “diffuse white”), are:

  • Average = higher than 0.15
  • Dim = 0 to 0.15
  • Dark = 0

Room 1.1. An HDR Hero Grade via ACES 2.0’s ODT “P3-D65 ST2084 1000 nit” (Output.Academy.P3-D65_1000nit_in_P3-D65_ST2084):

  • Grey Cards (float 0.18) displayed and measured by Resolve’s 10-bit Picker = 337/1023 followed by the monitor’s PQ-EOTF ≈ 14 nits
  • “Diffuse White objects” (float 1.00), 526/1023 followed by the monitor’s PQ-EOTF ≈ 106 nits
  • Surround ratio: 5 nits surround / 106.2 nits displayed ≈ only 0.05
    • Rather near “Dark” surround ratio, or still “Dim”?

Room 1.2. A subsequent SDR-trim via ACES 2.0’s ODT “Rec.709 BT.1886” (Output.Academy.Rec709-D65_100nit_in_Rec709-D65_BT1886):

  • Grey Cards (float 0.18) displayed and measured by Resolve’s 10-bit Picker = (392/1023)^2.4 at 100 nits monitor calibration ≈ 10 nits
  • “Diffuse White objects” (float 1.00), (739/1023)^2,4 at 100 nits monitor calibration ≈ 46 nits
  • Surround ratio: 5 nits surround / 45.8 nits displayed ≈ 0.11
    • Clearly “Dim” surround ratio, but in the very same room as above.

Room 2. A subsequent Cinema-trim via ACES 2.0’s ODT “P3-D65” (Output.Academy.P3-D65_48nit_in_P3-D65_Gamma2pt6):

  • Grey Cards (float 0.18) displayed and measured by Resolve’s 10-bit Picker = (422/1023)^2.6 at 48 nits projector calibration ≈ 5 nits
  • “Diffuse White objects” (float 1.00), (757/1023)^2,6 at 48 nits projector calibration ≈ 22 nits
  • Surround ratio: 0.01 nits ambient* / 21.94 nits displayed ≈ 0.00046 ≈ 0.00
    • Clearly close to “Dark” surround ratio (and even closer to “Dark” if lowering the ambient light).
    • * RP 431-2: “For Review Rooms, the ambient light level reflected by the screen should be less than 0.01 cd/m2”.

The purpose of ODTs:

In my mind: A today non-existent reference monitor able to fully display an ACES 2065-1 based Non-Graded Archival Master (NAM) will ideally reproduce the light, that hits the sensor-plane behind freely chosen lenses and creative filters, captured as the “reality”. I.e., The scene-referred “reality”, where the scene lighting, the choice of lenses and creative filters are the only creative tools.
The largest colour volume currently standardised for display, “10 000 nits PQ” based, must in the ODT add tone mapping, gamut compression, colour appearance compensations et cetera, plus a viewing environment compensation, to reproduce the perception of the “reality” on set, but not add any subjective creativity.
When going down further to smaller colour volumes to be displayed, we must still target the perception of the “reality” on-set in each ODT and clearly connect each ODT to a specific viewing environment.

The current situation:

When comparing the, contrast wise, very different “Rec.709 BT.1886” ODTs in ACES 1.3 versus 2.0, the Grey Gard stays displayed near 10 nits in both versions, but a diffuse white object falls from 62 nits (in “contrasty” ACES 1.3) down to 46 nits (in ACES 2.0). I cannot help wondering:

  • Question 2: Which one of those two ODT-versions reproduces the most accurate perception of the “reality” on-set?
  • Question 3: If subjective “creative” modifications, not only strict pure colour appearance compensations to preserve "reality on-set, are introduced in ACES 2.0’s ODTS, is also then viewing environment compensation deliberately/indirectly baked-in already in ACES 2.0’s ODT?

HDR broadcasting/streaming:

ACES 2.0’s ODT “P3-D65 ST2084 1000 nit” displays diffuse white at 106 nits. The colourists, at the company I work for, are indeed comfortable with that initial alignment of mid grey at 14 nits, it initially feels like a comfortable luminance level and nobody has experienced any need to start up the grading session by “offsetting” the whole timeline overall.

However, regarding Rec.2100, the recommended alignment levels are the high 26 nits regarding Grey Cards and the high 203 nits regarding diffuse white objects. The colourists (where I work) believe such high levels feel “too much” when they in Resolve’s ACES 1.3 are able, purely as a test, alter the mid grey alignment level with a slider in the GUI.

As we all remember the “loudness war”, between commercials and drama-productions, regarding dialogue in audio back in the days:

  • Question 4: Are there any up-to-date studies worldwide showing that camera control operators, regarding HDR live broadcasting, really are targeting such high mid grey exposure levels as recommended in Rec.2100? Or, are they year 2026, in true practice, acting more relaxed? I.e., should the ITU reconsider, and broadcast delivery specifications, and QC, be revised?
  • Question 5: Back in 2018, MovieLabs published “Best Practices for Mapping BT.709 Content to HDR10 for Consumer Distribution” and recommended to scale (gain) legacy SDR-content by x2 (100 nits stretched up to 203 nits), but is such a strong compensation, today in true practice, actually used in HDR broadcasting/streaming - or has a more relaxed SDR to HDR compensation developed over the years?

In advance thanks for all answers, comments, and ideas!

Lars HAGLUND
Sweden

2

Hi Lars,

Is the overall theme of your post about a getting a faithful representation of a scene in a given env.? Or, are you proposing a study?

Cheers!

3

The idea to study viewing condition compensation is a meaningful endeavour.

However the notation of “depicting reality” and “unbiased view” is the wrong direction I believe.

And we have very established heuristics how to convert from one viewing condition to the other within the traditional dynamic ranges - dark, dim and average.

4

You got to my next step before me @daniele! Thanks for the reply!

5

Thanks Charles, Daniele, all,

The overall theme is future ODT versions. Then the following sub-themes:

  • Should ODTs include baked-in subjective “look” impact, or should ODTs strive to act as neutral as possible, depicting “reality” (see my original text for my definition of “reality”) plus be clearly connected to a specific viewing environment compensation (each), hence put all subjective “look” impact into the ACEScct based grading?
    • Please correct me, but it could be (based on the large contrast difference between ACES 1.3’s and ACES 2.0’s “Rec.709 BT.1886” ODTs) that a larger amount of creative impact became included in the ACES 2.0. version?
  • Should Viewing Environment Compensation, likely available in Colour Appearance Models, be used in ODTs?
    • Please correct me, another thread here suggests it never became used when developing ACES 2.0?
  • It was more than 10 years ago, the exposure of Mid Grey and Diffuse White were discussed in the ITU during the development of Rec.2100. So, is ACES’ (IMHO comfortable) Mid Grey alignment level of HDR at 14 nits appreciated by all those that receive our HDR deliverables, and have general- and sports-programming in HDR come down to our mid grey luminance level during these 10 years?

Kindly, Lars

6

Hello, didn´t the OT VWG of ACES 2.0 proved that this goal was some kind of fantasy ?

7

Thanks Christophe,

I am too wondering if a “top to down” concept, i.e, “optical reality as the reference, down to more and more limited display via a Colour Appearance Model” was judged as a “fantasy”?

Please bear with me, when dreaming in more detail:

  1. Imagine that we buy a container.
  2. Inside the container we paint all walls with “Musou Black” to create a “camera obscura”.
  3. We mount a traditional, but in physical length extended, optical view finder (with the possibility to add optical ND filters) to the non-dominant eye of picky colourists (not humans in general), dressed in “Musou Black” clothing, helping the study.
  4. We build a dividing wall between non-dominant and dominant eye. The dominant eye is watching (and separately adapting to) a state-o-the art display (e.g., better than Dolby’s Pulsar) with adjustable surround behind.
  5. The camera is an ARRI Alexa 35 because of it dynamic range and is framing the same scene-area for the dominant eye as the optical viewfinder simultaneously is for the non-dominant eye.
  6. The colourists use physical sliders to alter the parameters of the (CIE) Colour Appearance Model (acting as the ACES ODT).
  7. The container (or “rig” + tent of “Musou Black” textile) is moved to several appropriate scene environments, both indoors (including lighting that causes out-of-gamut-issues) and outdoors.

All step by step built-up colour knowledge since the 1960’s would be confirmed “live”. The ACES community would get conceptually true ODTs where the baked-in RRTs are established “live”. In addition, CIE would get notices about issues where the CAM could be improved.

Maybe this dream is already, though indirectly, fulfilled in the current ODTs? I do not know.

Kindly, Lars

(Note: Not just the ACES community, CIE, Rochester Institute of Technology et cetera may be interested in such a moveable test set-up and effort, but also a commercial company like Dolby, that after all ACES ODT tests, might want to improve their Dolby Vision analysing and initial tone mapping further?)

8

I agree with your statement Christophe. Subjectivity is always at play. I feel that the best we can do is agree on a set of base DRT as a best starting point. Which was the point of VWG. What we have is a “best compromise” for the time being. That being said: the Cobalt Standard Spectral HDR Lightroom setup (dcp) by Cobalt Image I use with my own camera is the closest I have ever been to having a drt that looks like the scene (to my own taste). But, in all cases the imaging chain is impossibly objective (in the current state of things).