ADOBE AFTER EFFECTS: Forum Expressions Tutorials Creative Cloud

Re: Color space problems again

COW Forums : Adobe After Effects

Respond to this post   •   Return to posts index   •   Read entire thread

Andrew Somers
Re: Color space problems again
on Oct 17, 2018 at 11:52:40 am

Hi Pierre,

Jerry's issue has to do with his workflow and encoders, and if I recall some interpretation settings, and I believe he is using color management.
Pierre said: Although sRGB and rec709 are close enough,

Rec709 and sRGB use identical color primaries and white point. Other than the transfer curve, they are not just "close enough", the primaries are interchangeably identical (if working in a linearized space, there is no difference in choosing sRGB or Rec709 as far as the working space is concerned — the only difference is in the output module when rendering to a gamma-encoded file format).
Pierre said: gamma 2.2 and gamma 2.4 is not

Just to be pedantic: sRGB (file encoding) is not gamma 2.2, but a parametric curve. Some (badly made) sRGB ICC profiles may use a simple 2.2 gamma curve, but the correct transfer curve is actually a gamma curve of 2.4, offset by a linear section near black, that results in approximating a 2.2 curve. The linearized section is to allow the curve to be inverted using integer math (computationally less expensive than float).

Rec709 is never gamma 2.4,( though there is much confusion on this point.) Rec709 defines the source/file encoding and Rec1886 defines the DISPLAY, and Rec2035 defines the VIEWING environment.

Rec709 is *also* a parametric curve that approximates a 1.9-2.0 gamma curve. Here the linearized section was intended to help with noise in the black areas of camera signals of the day (remember the spec was developed circa 1990, well before digital HDTV emerged).
Rec709 CURVE

But these show the curve as encoded in the signal or fileNOT in the DISPLAY itself. The Rec709 curve is "source referred" and even though the sRGB curve is called "display referred",NEITHER curve defines the actual display gamma of the monitor in any particular instance.

The sRGB curve for the file encoding is parametric, but the actual monitor spec is a simple gamma 2.2 curve.

Using log intensity in this graph to make it more apparent, note the difference between the sRGB curve (file encoding) and a simple 2.2 gamma curve (Idealized display).

But in addition, displays are very user adjustable. Some even have a control labeled gamma, but most or all have BRIGHTNESS (gain) and CONTRAST (lightness/luminance) – and unless a hardware calibration and profiling is done, they could have been set to an "equivalent" gamma from 2.1 to 2.5 or more.

(Also, remember that legacy CRT displays were typically a gamma of 2.4 — see "Digital Video and HD" by Poynton)

In the case of Rec709, the 1.9~2.0 signal/file gamma is fed into a monitor with an EOTF display gamma of 2.4, per Rec1886. Rec709 defines ONLY the SIGNAL or FILE gamma. Rec1886 defines the MONITOR gamma.

Side Note:
Both sRGB and Rec709 have a "system gamma" that increases the gamma by the time it is displayed. For sRGB it's a power of about 1.1, and it's a power of about 1.2 for Rec709.

Side Side Note:
If you are working in a color managed system, then the actual physical gamma of the display is less relevant or simply irrelevant. Assuming the display is profiled with a hardware colorimeter, the effective system gamma will be whatever the CMM adjusts it to, based on various settings.

Pierre said: Your computer monitor is likely sRGB gamma 2.2 and your TV is likely rec709 gamma 2.4.

The monitor and the TV are likely all over the place. Both have easy to access user controls that can shift the effective or actual gamma all over as a typical user adjusts for "what they like" for any given environment. Unless they are adjusted with hardware (a colorimeter), there is nothing to enforce a 2.2 or 2.4 gamma.

And in fact, unless the monitor is in a calibrated viewing environment (such as Rec2035), then the specified gamma of sRGB (or Rec1886) may in fact be wrong in terms of creating the same perceptual uniformity. In fact the reason for the increased gamma in the system is because the expected viewing environment is dimmer than the environment while the image is being captured with a camera. The brightness of the viewing environment critically changes the perception of the displayed image.

Rec 709 related display and room specs:
Gamma: 1.9~2.0 (parametric curve)
White point: D65

Rec 2035 Viewing Environment:
Room Illumination: 10 LUX
Ambient white point: D65
BG to Screen luminance ratio: 10% of screen reference white

Rec 1886 Display Qualities:
Gamma: 2.4
Reference White Luminance: 100 cd/m2 (meaning m squared)
White point: D65

sRGB related display and room specs:
Gamma: ~2.2 (parametric curve)
White point: D65

sRGB Viewing Environment:
Room Illumination (Encoding): 64 LUX
Room Illumination (Typical): 200 LUX
Ambient white point: D50
BG to Screen luminance (surround reflectance): 20% of screen reference white

sRGB Display Qualities:
Gamma: 2.2
Reference White Luminance: 80 cd/m2 (meaning m squared)
White point: D65

Notice the substantial difference in ambient light for the Rec2035 view environment (modeled after a dark living room) vs the sRGB view environment (modeled after a typical office environment).

The point to this long winded post is simply to indicate the importance that the FILE encoding is NOT intended to be the same as the DISPLAY'S GAMMA, as the overall system gamma is designed to increase the gamma from end to end.

Thus a Rec709 file should never be encoded as gamma 1/2.4, but should be the parametric curve closer to 1/2.0, thoiugh it should be noted that the viewing of the working space SHOULD be setup (such as with a view LUT) to display that system gamma gain.

Andrew Somers
VFX & Title Supervisor

Posts IndexRead Thread 

Current Message Thread:

© 2019 All Rights Reserved