Notice of Meeting - Output Transforms Architecture VWG - August 30th, 2023

Output Transforms Architecture VWG Meeting

Wednesday, August 30th, 2023
1pm PDT Add to calendar (.ics)

Dropbox Paper homepage for this group:

NOTE: We will be using Zoom for this meeting.

You may join via computer/smartphone (preferred) which will allow you to see any presentations or documents that are shared, or you can join using a telephone which will be an audio only experience.

Please note that meetings are recorded and transcribed and open to the public. By participating you are agreeing to the ACESCentral Virtual Working Group Participation Guidelines.

Audio + Video
Topic: ACES Virtual Working Group Meetings
Time: This is a recurring meeting Meet anytime

Join Zoom Meeting

Or follow this link:

Meeting ID: 993 4048 4145
Passcode: 972426
One tap mobile
+16699006833,99340484145#,*972426# US (San Jose)
+16694449171,99340484145#,*972426# US

Meeting ID: 993 4048 4145
Passcode: 972426
Find your local number: Zoom International Dial-in Numbers - Zoom

The recording and notes from meeting #116 are now available.

Might be worth mentioning that post the meeting I was also able to recreate the behaviour we showed at the end of the meeting with near violet using both CIECAM 16 as well as a pure implementation of Luke’s model, suggesting that the root of the problem is unlikely due to any of our modifications. But likely lives in the common portions (unfortunately this is most of the models)


1 Like

I think what you are seeing are the differences in the J and Y weightings. You cannot tonemap Y and expect to get the same result in J or any other signal once you plug in the original other two components.

Also a question about the image you are using:
How many watts of energy do you need to put into a room to get a measurable or perceivable visual stimulus if the wavelength is close to ultra violett laser. It would be quite dangerous for the eye to look at such a scene. I guess this is also why the system gets unstable in this area.
Don’t get me wrong, it is a very nice test to stress a system, but is the input data “visually plausible”.

1 Like

I have no idea if this is specifically related to the “near violet” problem, but it may not be completely unrelated…

As I mentioned in last meeting here are diagrams of the triangles created by LMS Primaries for CAM16/CAT16 and Pekka’s modified matrix.

All values outside the triangles will produce negative values.

Yxy can be used to show the boundary but CIE u’ v’ is required to show closed triangle.

CAM 16

Pekka Matrix

1 Like

Would be interesting to see how the spectra near ultra violet behaves when transformed through the corresponding Spectral response curves for CAT16?

The results should be equivalent, but might still be good to visualize the interaction.

I don’t have the data for those curves, it does exist, but was it ever published?

I don’t think the primaries have anything to do with that.
But wow I thought you guys just nudged the proxy LMS space a tiny bit. But from the plots it looks that we are talking about two very different gamut definitions.

It is probably better to assess the effect of the primaries rather than the primaries themselves:

The intent was to move the blue focal point outside of AP0 (and straighten the lines): ZCAM for Nuke - #231 by Thomas_Mansencal



the Blue/S numbers in Pekka’s current matrix are just a nudge from CAM16 but they get exaggerated in the Yu’v’ diagram.

Here is how they look in Yxy, which does work as a closed triangle, this appears to be close, but not exactly the same as Thomas’ diagram, since Pekka has made further tweaks.

I also don’t think the LMS is the is the source of the problem, or at least not the sole source, but I did find it curious how the CAM16 triangle almost touches edge of boundary at 380nm.

I also want to point out that my diagram does not fully agree with diagrams in CAM16 paper for some reason…

And my adjustments were to try and bring the green and yellow hue lines back closer to what the original primaries gave (with the LMS compress mode enabled).

Watched the recording of the meeting again. Also glad many with much more experience than I are on this.

It still seems to me that the ‘tilt’ along the magenta-green direction, between the calculated 100 nit edge of the ‘bowl’ (contained cusp call it) and the plane of 100 nit is directly related to the wildly curving up in the magenta area. Might some power, exponential or hyperbolic be doing this? But then, what do I know as I am still impressed by an arc-sine of a sine being able to produce a sharp triangle wave. However this is not magic nor does it seem natural behavior so there must be something in the math making this happen.

It does seem to happen early on in the code and compression was said to be off. Maybe comment out one line at a time to find where it comes from. I understand there are hundreds of lines of code, but many won’t be capable of contributing to this issue. Some math has to be making this non-linear offset.