Short answer
Scene-referred values describe the light that was in front of the camera; display-referred values describe the light a particular kind of screen should emit. A camera log file and an ACES frame are scene-referred, and a Rec. 709 master or a PQ deliverable is display-referred. The step between them, an output transform, is where a scene's range is squeezed into a display's, and keeping the work on the scene side of that step is what lets one grade serve cinema, SDR television and HDR.
A pixel value of 0.18 can mean 'this surface reflected 18 per cent of the light' or 'drive the screen to 18 per cent'. The first is a statement about the scene and has no upper limit: a lamp in shot may be a thousand times brighter than a grey card. The second is a statement about a display and stops at that display's white. Neither is more correct, but operations mean different things in each. Doubling a scene-referred value is a one-stop exposure change. Doubling a display-referred value is a brightness change that runs into the ceiling.
Camera log encodings are scene-referred: the ARRI specification describes LogC4 in those terms. The ACES documentation describes a framework centred on a standardised scene-referred encoding. Delivery formats are display-referred, with one instructive exception inside the HDR standard itself. ITU-R BT.2100 defines PQ by a function that maps the signal to display light, in absolute terms, and defines HLG by a function that maps relative scene light to the signal, leaving each display to apply its own rendering. The table below sorts the encodings Colourwise holds records for.
Going from scene to display is not a format conversion. A sunlit scene spans a far greater range than any screen, so the transform has to compress highlights, set contrast through the middle and decide what becomes of colours too saturated for the display. RED's pipeline note is candid that its output transform contains technical controls, a tone-map curve and a highlight roll-off, that impart an aesthetic. ACES 2 does the equivalent with a tone scale, chroma compression and gamut compression. Whichever is used, it is a large part of the look.
If corrections are made before the output transform, the same graded data can be rendered for a dim cinema screen, a 100 cd/m² broadcast monitor and a 1,000 cd/m² HDR display by changing only the last step. RED calls this monitor independence and advises that all grading happen before the output transform. Grade after it, on display-referred values, and the highlights have already been compressed for one display; there is nothing left above white to bring back for another. This is the practical reason the order of operations matters.
American Cinematographer's account of Roma follows the chain end to end. The footage arrived as linear files in the camera maker's wide gamut. It was converted to a log encoding as it entered the colour corrector, and the creative grading was done in that log state. A custom output transform at the end supplied the overall look and also removed the colour, so the black-and-white picture was a rendering decision made last. Because the colour was still present upstream, the colourist could use it to select parts of a monochrome image.
| Encoding | Kind | Refers to | Note |
|---|---|---|---|
| Cineon | Film-scan encoding | Scene-referred | Kodak's handbook dates the system; the encoding's code values come from Kodak documents that were not read, so none are given. |
| ARRI LogC3 with ARRI Wide Gamut 3 | Camera log encoding | Scene-referred | The curve's parameters vary with the camera's exposure index setting. |
| ARRI LogC4 with ARRI Wide Gamut 4 | Camera log encoding | Scene-referred | Not interchangeable with LogC3: a LogC3 lookup table applied to LogC4 footage is wrong. |
| Sony S-Log3 with S-Gamut3 or S-Gamut3.Cine | Camera log encoding | Scene-referred | Taken from Sony's user guide; the technical summary that defines the curve could not be read. |
| RED Log3G10 with REDWideGamutRGB | Camera log encoding | Scene-referred | Only the pipeline note was read in full; no curve values are taken from the white paper. |
| ACES2065-1 | Working space | Scene-referred | An exchange encoding; grading and rendering are done in the two encodings below. |
| ACEScg | Working space | Scene-referred | A working space, not a delivery or archive format. |
| ACEScct | Working space | Scene-referred | Transient by design: the specification says it is not for interchange or archive. |
| ACES 2 output transform | Transform | Display-referred | An ACES 1 and an ACES 2 rendering of the same grade differ; a project has to hold one version throughout. |
| ITU-R BT.709 | Delivery standard | Display-referred | Defines the camera curve, not the display's; that is BT.1886. |
| ITU-R BT.1886 | Monitoring | Display-referred | Assumes a dim reference room; the same file on a bright desktop reads darker and flatter. |
| ITU-R BT.2020 | Delivery standard | Display-referred | Used mostly as a container: masters are graded inside a smaller gamut and carried in it. |
| ITU-R BT.2100 (PQ and HLG) | HDR format | Display-referred | HLG is scene-referred where PQ is display-referred; the single label here is for PQ. |
| SMPTE ST 2084 | HDR format | Display-referred | Only the publication record was confirmed; the curve is described from BT.2100. |
| HDR10 | HDR format | Display-referred | Summarised from a reference work because the association's definition could not be read. Static metadata cannot describe a change from scene to scene. |
| Dolby Vision | HDR format | Display-referred | The owner's 2019 guide; two content-mapping versions exist and their trims differ. |
| DCI-P3 | Cinema standard | Display-referred | Not the same as Display P3, which pairs the same primaries with D65 and the sRGB curve. |
| Digital Cinema Package | Cinema standard | Display-referred | Described from the Library of Congress summary; the specification's colour clauses were not read. |
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
StandardStrong evidence
ITU-R BT.2100-3 specifies the PQ system by a reference function that maps the non-linear signal to display light, and the HLG system by a reference function that maps relative scene linear light to the signal.
FactStrong evidence
The ACES documentation describes the system as a framework centred on a standardised scene-referred colour encoding, with input transforms converting camera data into it and output transforms preparing it for a display.
Source: ACES Documentation: system overview, encodings and output transforms
FactStrong evidence
RED's IPP2 note places grading between raw development to a single log encoding and an output transform, states that the output transform is what gives monitor independence, and advises that all grading take place before it.
Source: IPP2 Output Transforms (RED document 915-0201 Rev A)
FactModerate evidence
Roma was graded in a log encoding converted from linear camera files, with a custom output transform applied last that supplied both the overall look and the conversion to black-and-white.
Source: Roma: Memories of Mexico (Mark Dillon, American Cinematographer, 7 January 2019)
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.