Re: [w3ctag/design-reviews] [wg/png] PNG Working Group Charter (Issue #1143)

svgeesus left a comment (w3ctag/design-reviews#1143)

> could you explain to me what "improving HDR & Standard Dynamic Range (SDR) interoperability" means in practice?

Sure. If we look for a moment at the [PNG Third Edition Explainer](https://github.com/w3c/png/blob/main/Third_Edition_Explainer.md), the section on [HDR Support](https://github.com/w3c/png/blob/main/Third_Edition_Explainer.md#hdr-support) calls out three features that were added in that edition:

1. [Labeling HDR content (`cICP`)](https://github.com/w3c/png/blob/main/Third_Edition_Explainer.md#labelling-hdr-content). This says what colorspace the image uses. This is well implemented and [tested](https://w3c.github.io/png/Implementation_Report_3e/#cicp) because you can make test files that are the same actual color but in different colorspaces, and verify that the produce the same color on screen.
2. [Mastering Color Volume (`mDCV`](https://github.com/w3c/png/blob/main/Third_Edition_Explainer.md#mastering-color-volume)). This says what colors the monitor used to produce the content could display, which will often be a subset of the image colorspace. A common case is a BT.2100 image mastered on a P3D65 monitor. This is passive metadata: the information may be helpful for processing (such as deciding whether to gamut map or to simply clip) but we don't say what to actually do with it.
3. [Content Luminance Levels (`cLLI`)](https://github.com/w3c/png/blob/main/Third_Edition_Explainer.md#content-luminance-levels). This gives statistical data on the average luminance level of the brightest frame, and the luminance of the brightest pixel. This can be used to tone map an HDR image for a display with less capable HDR. PNG recommends a way to calculate this, but again this is passive metadata; we don't say what to do with it.


Those three items are needed for HDR10 compatibility, but only one is actually testable and so implementations differ in how they handle it. There are industry specifications (from SMPTE) that define tone mapping but we do not require them to be used (and, until PNG WG successfully negotiated with SMPTE, these specifications were not freely available). We would like to improve on this situation.

In addition, ISO 22028-5 (which has not been released yet, but we have been working with ISO TC 42 who develop  this spec) also requires [Diffuse white luminance metadata](https://github.com/w3c/png/issues/390) which we are considering adding to PNG Fourth Edition.

Also, another technology from TC 42, [gain maps](https://github.com/w3c/png/issues/380), is a way to effectively provide two images (one HDR with a high headroom, another SDR with zero headroom) and to smoothly interpolate between them to match the actual amount of headroom available. We are considering adding that to PNG Fourth Edition as well. Gain maps are implemented in, for example, Adobe Photoshop, for JPEG images. We would like a compatible gain map for PNG. (The ISO spec is also not released yet, and is not freely available, but we are [working with them](https://github.com/w3c/png/issues/366) and have seen drafts).

-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3ctag/design-reviews/issues/1143#issuecomment-3233007799
You are receiving this because you are subscribed to this thread.

Message ID: <w3ctag/design-reviews/issues/1143/3233007799@github.com>