Re: [w3ctag/design-reviews] WG New Spec: HDR on the web (CSS, Canvas, WebGL, WebGPU) (Issue #1194)

ccameron-chromium left a comment (w3ctag/design-reviews#1194)

Thank you for the review!!

> However, it's possible that the correct design is, for example, to use `linear` in most places for consistency with color values, but to use log2 in places that are closely tied to image formats that already use log2.

This is the direction that we'll try to aim for. In terms of landing features, we'll be first adding support for the places that are perforce `"linearHdrHeadroom"`. The specs that align closely with log2 (ISO 21496-1 and ST 2094-50) will be in the next phase. We'll likely re-run them by you in that phase.

> We noticed that https://github.com/w3c/ColorWeb-CG/blob/main/hdr-big-picture.md#css-hdr-colors says "The [hdr-color](https://drafts.csswg.org/css-color-hdr/#funcdef-hdr-color) functional ... allows an CSS color to color match a pixel in an HDR image at all HDR headrooms." However, [color-hdr()](https://drafts.csswg.org/css-color-hdr/#funcdef-hdr-color) appears to only allow geometric-mean interpolation (or whatever the right term is) between specified headrooms, while the SMPTE ST 2086 tone mapping seems more flexible.

Yes, good catch. Indeed:

- `color-hdr()` allows interpolation between 2 endpoints
- ISO 21496-1 allows interpolation between 2 endpoints, with an $\epsilon$ offset
- ST 2094-50 allows 5 endpoints, with no $\epsilon$ offset.

I think we're going to start with `color-hdr()` syntax being the intersection between ST 2094-50 and ISO 21496-1. So, 2 endpoints, no $\epsilon$. But we will want to design the syntax so that we can add all the bells and whistles if we really want to.

While the specs formulate things in very complicated ways because of how they store things, they all boil down to geometric mean when the math gets run.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3ctag/design-reviews/issues/1194#issuecomment-4378235889
You are receiving this because you are subscribed to this thread.

Message ID: <w3ctag/design-reviews/issues/1194/4378235889@github.com>