Add non-normative Appendix to cover SDR compositing.

[jmccrossan-fox]

Appendix Q (non-normative) covers High Dynamic Range Compositing. Consider adding an additional non-normative appendix to cover SDR compositing of sRGB subtitle code values (mapping from sRGB Gamma 2.2 / Peak White 80nits) where the presentation output is BT.2020 or BT.709 (Gamma 2.4 / Peak White 100nits).

[nigelmegitt]

@jmccrossan-fox Thank you for raising this issue.

It seems that this should be a problem that has been solved before. Are you able to suggest any existing reference documentation about the mapping of sRGB into BT.709, say, that we could make use of in resolving this?

[jmccrossan-fox]

Hi @nigelmegitt thanks for reviewing!

Indeed sRGB->BT.709 mapping is well understood (although unfortunately many consumer electronic devices simply treat sRGB as BT.709 when presenting over BT.709 video). We will work on preparing a contribution with the objective to use an existing reference (e.g. ITU) if one exists.

[cconcolato]

@palemieux the need for this clarification came in part from IMF discussions. Should it be in scope of TTML2 2nd Ed?

[palemieux]

@cconcolato Let me check.

[palemieux]

Perhaps @dkneeland can provide text.

[dkneeland]

Here's my recommendation:

1. Let (r, g, b) be a full-range sRGB pixel with an opacity of 1
2. Invert the 8-bit full-range quantization
   (r, g, b) / 255 -> (r, g, b)
3. Linearize using the sRGB reference display EOTF
   (r^2.2, g^2.2, b^2.2) -> (r, g, b)
4. Apply inverse rec. 709 reference display EOTF
   (r^(1/2.4), g^(1/2.4), b^(1/2.4)) -> (r, g, b)
5. Apply 8-bit full-range quantization
   (r, g, b) * 255 -> (r, g, b)

[css-meeting-bot]

The Timed Text Working Group just discussed Add non-normative Appendix to cover SDR compositing. ttml2#1119, and agreed to the following:

• SUMMARY: TTWG Thanks @dkneeland for this contribution, and is considering when it can be added to the specification, within TTWG publication timelines.

The full IRC log of that discussion

<nigel> Topic: Add non-normative Appendix to cover SDR compositing. ttml2#1119
<nigel> github: https://github.com//issues/1119
<nigel> Nigel: Do we need to agree that the EOTF linearization uses gamma 2.2 and the inverse 709 EOTF uses 2.4 and there's
<nigel> .. no change to the primaries? Do we need to worry about number ranges?
<nigel> Pierre: I don't think we need to worry about number ranges, because both 709 and sRGB are relative luminance systems
<nigel> .. so I think we can use the full range. Both 709 quantisation schemes are in use. We don't need to cover narrow range
<nigel> .. 709 - this example covers full range 709.
<nigel> .. Full range 709 is used in practice.
<nigel> Nigel: Thanks for that.
<nigel> Pierre: It depends on the application.
<nigel> Nigel: Great, so as long as it's clear that we're talking about full range 709 we're ok
<nigel> Pierre: Yes, we should change the title to avoid people getting confused.
<nigel> Nigel: And those gammas are uncontroversial?
<nigel> Pierre: As far as I know, yes. If you don't apply this conversion then colours won't match in practice.
<nigel> Nigel: OK so the change essentially looks good, then the question moves to when we should implement it, should we
<nigel> .. attempt to shoehorn it into TTML2 2nd Ed or wait until 3rd Ed?
<nigel> .. I note that this is an informative section.
<nigel> Glenn: There's a bigger question that Pierre asked which is have we stopped making changes to 2nd Ed other than
<nigel> .. typos basically, which is more general than this particular issue. I think we should ask the group to make a decision
<nigel> .. on this.
<nigel> SUMMARY: TTWG Thanks @dkneeland for this contribution, and is considering when it can be added to the specification, within TTWG publication timelines.