Some reflections and experiments

I’ve had a little play with this is SDR and HDR now.

Compared with the current v12 of the stock ZCAM DRT (Candidate C), the SDR rendering is desaturated, whilst the HDR one is more saturated, which to my eye reduces the quality of the match between the two renderings. Subjectivly anyway.

I feel like in SDR it’s a win for images that have data of questionable quality at the top end, hiding some sins. But I’m not sure if that’s a net gain if it means it’s harder to hit bright saturated corners of the display when you do have good data up there. Killing the saturation of bad highlight data is something eaisly done in a grade op, but the inverse is not true if it’s being sent to grey by the DRT.

These are my first impressions anyway.

Just to verify, did you test against the OkishDRT or the ZCAM DRT with the derivative driven desaturation?

Do you find that this behaves similarly to per-channel transform like rgbDT in HDR or is it worse? The default settings I have there push things a bit and I wouldn’t be surprised if HDR would have to have its own settings.

This to me too is still something to explore, to try and still have the ability to hit those corners with this technique. That is one thing I like about the current approach in ZCAM DRT.

It occurred to me that the derivative used to do the desaturation is effectively just a mask, so it doesn’t have to be the exact derivative. It can be modified, or based on some other tonescale or be in fact any compatible curve. I tested with Daniele’s original MM tonescale and used its derivative as the path to white (but not as tonescale) and it works fine. The flare parameter can be used to precisely affect the saturation in the shadows, contrast changes the saturation in the mid tones and the peak luminance can be used to change how the top end desaturates. The only thing it can’t be used to adjust is the shape of the shoulder, which would be nice extra adjustment to control the top end saturation more precisely. Having parameters separate from the tone scale parameters probably makes sense.

It really does show how subjective this all is. I have always found the HDR rendering of ZCAM a bit desaturated compared to the SDR to my eye. So I personally prefer the behaviour of @priikone’s rendering.

(Caveat, I don’t have a reference monitor, so am viewing on an LG OLED TV)

I was thinking about this again today. Perhaps the Okish HDR is more saturated than the SDR, but my brain is anticipating the HK effect, so a brighter image being more saturated feels “correct” to me, and desaturating to compensate feels “wrong”.

The plot thickens!

1 Like

There is no right or wrong here, just preference and viewing experience.
My own pivot about this topic changed several times over the course of the last few years.
Mainly because of the pragmatic pros and cons that come with the different approaches.

I would not go down that rapid hole and try to explain your current preference with some half-understood (in general) colour appearance phenomena because they do not apply to complex stimuli. Further modelling vision in a per-pixel transfer is doomed anyway.

I think it is clear that objectiv reasoning does not bring us any further. There will be no objective correct solution to this problem. If we agree on that, the task becomes much simpler.


Daniele is right. At some point, things become a matter of preference and some content works better than other with one DRT or another depending on how it was look-deved. As an example, when we tried OpenDRT at Larian, we got two different kind of reactions: artists who didn’t compensate for ACES 1 hue skews liked it while others who had specifically done colour picking in order to compensate for ACES 1 hue skews (or were exploiting the hue skews for another purpose) hated it because their content now looked all wrong. In the end, we removed it because of both the license change and because the modified version of v11 Candidate C combined with the LMT I made pleased everyone and I was able to close the JIRA.

As a supporting point to the SDR vs HDR saturation, this is where interactive media has an advantage : we can add calibration screens to allow users to tweak saturation, contrast and brightness to their liking. Preaching for my domain here (is that how it’s said in english???) :slight_smile:


Perhaps I was wrong to attribute it specifically to the HK effect. But to my eyes, what the Okish DRT does when switching to HDR “feels more like what I expect”, for whatever reason. But for others (including my wife, looking at the same LG OLED as me) what ZCAM does in HDR “matches expectations better” and the OKish DRT “doesn’t feel right”.

1 Like

Interesting. While it may mostly be about aesthetic preference, a more objective goal can be to make the HDR and SDR feel similar and have similar level of saturation. And to make sure colorist can take the HDR to a different direction with more colourfulness if that’s what they want.

Sorry, I should have been specific, I was looking at: DRT_ZCAM_IzMh_v12_deriv