Re: [EXTERNAL] Re: Display on operating systems requiring tone-mapping

For the test image that is provided, could you add a version of the image
where the "out of range" pixels (those with signal >0.75) are set to pink.

(I'm working on a test application to do this for video shot on iPhone, but
didn't quite make it for the meeting)

On Thu, Oct 21, 2021 at 6:16 PM Seeger, Chris (NBCUniversal) <
Chris.Seeger@nbcuni.com> wrote:

> For a lot of commercial material, they expect the graphic white to be
> consistent.  This is why prefer this absolute mapping unless there is an
> ambient light sensor.  This also allows other conversions to be consistent
> for graphic white.
>
>
>
>
>
> *From: *Joe Drago <jdrago@netflix.com>
> *Date: *Thursday, October 21, 2021 at 6:14 PM
> *To: *Lars Borg <borg@adobe.com>
> *Cc: *Seeger, Chris (NBCUniversal) <Chris.Seeger@nbcuni.com>, Simon
> Thompson - NM <Simon.Thompson2@bbc.co.uk>, public-colorweb@w3.org <
> public-colorweb@w3.org>
> *Subject: *[EXTERNAL] Re: Display on operating systems requiring
> tone-mapping
>
> > Using the HLG equations and iterator or pre-determined look-up table
> find the HDR peak luminance and system gamma for the given, mandated SDR
> White level.
>
>
>
> FWIW, this is how I approached this when attempting to convert HLG into a
> PQ render target (for displaying on Windows), given a predefined/forced SDR
> level (such as the Windows SDR slider). I basically did a binary search to
> find which HLG peak luminance yielded the input SDR level at 0.75, and then
> I'd convert the image into that destination peak luminance. I was honestly
> surprised to not see anyone else doing this, as it has always felt to me
> like the anchor point for HLG was ensuring that all incoming contents' 0.75
> signal was a consistent luminance when compositing.
>
>
>
> If anyone is curious to see what such a search looks like, I have it here:
> https://github.com/joedrago/colorist/blob/master/lib/src/transform.c#L117
> <https://urldefense.com/v3/__https:/github.com/joedrago/colorist/blob/master/lib/src/transform.c*L117__;Iw!!PIZeeW5wscynRQ!-_PeodaRXi1cwrR_OA1_a2QAD2TVBZEEu5g1Nt11foTiLA4d6xybd2ClkcXY9MS0Bg$>
>
>
>
>
>
>
>
>
>
>
>
> On Thu, Oct 21, 2021 at 11:30 AM Lars Borg <borg@adobe.com> wrote:
>
> I had the same thought.
>
> We only want the peak to move. Not the gray card.
>
>
>
> *From: *Chris Seeger <Chris.Seeger@nbcuni.com>
> *Date: *Thursday, October 21, 2021 at 8:27 AM
> *To: *Simon Thompson - NM <Simon.Thompson2@bbc.co.uk>, "
> public-colorweb@w3.org" <public-colorweb@w3.org>
> *Subject: *Re: Display on operating systems requiring tone-mapping
> *Resent-From: *<public-colorweb@w3.org>
> *Resent-Date: *Thursday, October 21, 2021 at 8:26 AM
>
>
>
> I have one concern here related to graphic white.  Using this method,
> won’t it move graphic white around by pushing down the entire raster
> average brightness?
>
>
>
> Cheers,
>
> Chris
>
>
>
> *From: *Simon Thompson - NM <Simon.Thompson2@bbc.co.uk>
> *Date: *Thursday, October 21, 2021 at 2:10 PM
> *To: *public-colorweb@w3.org <public-colorweb@w3.org>
> *Subject: *[EXTERNAL] Display on operating systems requiring tone-mapping
>
> Hi all,
>
>
>
> At the last meeting there was some discussion on the application of
> tone-mapping on operating systems, particularly looking at what should be
> done when the SDR white level was mandated by the operating system.
>
>
>
> I’ve investigated and I think one possible solution could be:
>
> 1.       Using the HLG equations and iterator or pre-determined look-up
> table find the HDR peak luminance and system gamma for the given, mandated
> SDR White level.
>
> 2.       If the calculated peak luminance is higher than that of the
> target monitor:
>
> a.       Convert the image from the extended sRGB format to display light
> by converting to HLG and then applying the HLG EOTF with the peak luminance
> and system gamma calculated in step 1
>
> b.       Apply a tone mapping curve on the Y signal – we’ve had good
> results calculating a bézier curve or a logarithmic curve
>
> c.       Convert the image back to extended sRGB
>
>
>
> To reduce processing time, it may be necessary  to pre-compute some of the
> curves and apply the nearest match.
>
>
>
> In the following examples I use a bézier curve.  The images are HLG but
> even when viewed on an sRGB monitor, differences can be seen in the
> highlights – the lights, clouds and the posters.  The shadows and mid-tones
> are untouched.
>
>
>
> [image: A picture containing text, outdoor, way, road Description
> automatically generated][image: Chart, line chart Description
> automatically generated][image: A picture containing text, outdoor, way,
> road Description automatically generated][image: Chart Description
> automatically generated]
>
>
>
> I hope to join on Monday night, but am traveling and may be delayed.
>
>
>
> Best Regards
>
>
>
> Simon
>
>
>
>
>
> --
>
> *Simon Thompson MEng CEng MIET*
> Senior R&D Engineer
>
> *BBC Research and Development South Laboratory*
>
>
>
>